Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
1. TDREPORT generation 2. Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
Following are the details of the patch set:
Patch 1/3 -> Adds event notification IRQ support. Patch 2/3 -> Adds Quote generation support. Patch 3/3 -> Adds selftest support for Quote generation feature.
[1] https://cdrdv2.intel.com/v1/dl/getContent/726790, section titled "TDG.VP.VMCALL<GetQuote>". [2] https://lore.kernel.org/lkml/CAAYXXYxxs2zy_978GJDwKfX5Hud503gPc8=1kQ-+JwG_kA... [3] https://lore.kernel.org/lkml/a69faebb-11e8-b386-d591-dbd08330b008@linux.alib...
Kuppuswamy Sathyanarayanan (3): x86/tdx: Add TDX Guest event notify interrupt support virt: tdx-guest: Add Quote generation support selftests/tdx: Test GetQuote TDX attestation feature
Documentation/virt/coco/tdx-guest.rst | 11 ++ arch/x86/coco/tdx/tdx.c | 194 +++++++++++++++++++ arch/x86/include/asm/tdx.h | 8 + drivers/virt/coco/tdx-guest/tdx-guest.c | 175 ++++++++++++++++- include/uapi/linux/tdx-guest.h | 44 +++++ tools/testing/selftests/tdx/tdx_guest_test.c | 65 ++++++- 6 files changed, 490 insertions(+), 7 deletions(-)
Host-guest event notification via configured interrupt vector is useful in cases where a guest makes an asynchronous request and needs a callback from the host to indicate the completion or to let the host notify the guest about events like device removal. One usage example is, callback requirement of GetQuote asynchronous hypercall.
In TDX guest, SetupEventNotifyInterrupt hypercall can be used by the guest to specify which interrupt vector to use as an event-notify vector from the VMM. Details about the SetupEventNotifyInterrupt hypercall can be found in TDX Guest-Host Communication Interface (GHCI) Specification, section "VP.VMCALL<SetupEventNotifyInterrupt>".
As per design, VMM will post the event completion IRQ using the same CPU on which SetupEventNotifyInterrupt hypercall request is received. So allocate an IRQ vector from "x86_vector_domain", and set the CPU affinity of the IRQ vector to the CPU on which SetupEventNotifyInterrupt hypercall is made.
Add tdx_register_event_irq_cb()/tdx_unregister_event_irq_cb() interfaces to allow drivers to register/unregister event notification handlers.
Reviewed-by: Tony Luck tony.luck@intel.com Reviewed-by: Andi Kleen ak@linux.intel.com Reviewed-by: Mika Westerberg mika.westerberg@linux.intel.com Reviewed-by: Erdem Aktas erdemaktas@google.com Acked-by: Kirill A. Shutemov kirill.shutemov@linux.intel.com Acked-by: Wander Lairson Costa wander@redhat.com Signed-off-by: Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com ---
Changes since v1: * Used early_initcall() instead of arch_initcall() to trigger tdx_event_irq_init(). * Removed unused headers and included headers for spinlock and list explicitly. * Since during the early_initcall() only one CPU would be enabled, remove CPU locking logic (like using set_cpus_allowed_ptr() or get_cpu())
Changes since v2: * Fixed commit log and comments as per review suggestions. * Used __irq_domain_alloc_irqs() instead of irq_domain_alloc_irqs() to pass affinity mask. * Used smp_processor_id() instead of hardcoding CPU0 as per review suggestion.
arch/x86/coco/tdx/tdx.c | 159 +++++++++++++++++++++++++++++++++++++ arch/x86/include/asm/tdx.h | 6 ++ 2 files changed, 165 insertions(+)
diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c index e146b599260f..49d7e11066c1 100644 --- a/arch/x86/coco/tdx/tdx.c +++ b/arch/x86/coco/tdx/tdx.c @@ -7,12 +7,17 @@ #include <linux/cpufeature.h> #include <linux/export.h> #include <linux/io.h> +#include <linux/spinlock.h> +#include <linux/list.h> +#include <linux/interrupt.h> +#include <linux/irq.h> #include <asm/coco.h> #include <asm/tdx.h> #include <asm/vmx.h> #include <asm/insn.h> #include <asm/insn-eval.h> #include <asm/pgtable.h> +#include <asm/irqdomain.h>
/* TDX module Call Leaf IDs */ #define TDX_GET_INFO 1 @@ -27,6 +32,7 @@ /* TDX hypercall Leaf IDs */ #define TDVMCALL_MAP_GPA 0x10001 #define TDVMCALL_REPORT_FATAL_ERROR 0x10003 +#define TDVMCALL_SETUP_NOTIFY_INTR 0x10004
/* MMIO direction */ #define EPT_READ 0 @@ -51,6 +57,16 @@
#define TDREPORT_SUBTYPE_0 0
+struct event_irq_entry { + tdx_event_irq_cb_t handler; + void *data; + struct list_head head; +}; + +static int tdx_event_irq __ro_after_init; +static LIST_HEAD(event_irq_cb_list); +static DEFINE_SPINLOCK(event_irq_cb_lock); + /* * Wrapper for standard use of __tdx_hypercall with no output aside from * return code. @@ -873,3 +889,146 @@ void __init tdx_early_init(void)
pr_info("Guest detected\n"); } + +static irqreturn_t tdx_event_irq_handler(int irq, void *dev_id) +{ + struct event_irq_entry *entry; + + spin_lock(&event_irq_cb_lock); + list_for_each_entry(entry, &event_irq_cb_list, head) { + if (entry->handler) + entry->handler(entry->data); + } + spin_unlock(&event_irq_cb_lock); + + return IRQ_HANDLED; +} + +/** + * tdx_event_irq_init() - Register IRQ for event notification from the VMM to + * the TDX Guest. + * + * Use SetupEventNotifyInterrupt TDVMCALL to register the event notification + * IRQ with the VMM, which is used by the VMM to notify the TDX guest when + * needed, for instance, when VMM finishes the GetQuote request from the TDX + * guest. The VMM always notifies the TDX guest via the same CPU that calls + * the SetupEventNotifyInterrupt TDVMCALL. Allocate an IRQ/vector from the + * x86_vector_domain and pin it on the same CPU on which TDVMCALL is called. + * For simplicity, use early_initcall() to allow both IRQ allocation and + * TDVMCALL to use BSP. + */ +static int __init tdx_event_irq_init(void) +{ + struct irq_affinity_desc desc; + struct irq_alloc_info info; + struct irq_cfg *cfg; + int irq; + + if (!cpu_feature_enabled(X86_FEATURE_TDX_GUEST)) + return 0; + + init_irq_alloc_info(&info, NULL); + + cpumask_set_cpu(smp_processor_id(), &desc.mask); + + irq = __irq_domain_alloc_irqs(x86_vector_domain, -1, 1, + cpu_to_node(smp_processor_id()), &info, + false, &desc); + if (irq <= 0) { + pr_err("Event notification IRQ allocation failed %d\n", irq); + return -EIO; + } + + irq_set_handler(irq, handle_edge_irq); + + /* + * The IRQ cannot be migrated because VMM always notifies the TDX + * guest on the same CPU on which the SetupEventNotifyInterrupt + * TDVMCALL is called. Set the IRQ with IRQF_NOBALANCING to prevent + * its affinity from being changed. + */ + if (request_irq(irq, tdx_event_irq_handler, IRQF_NOBALANCING, + "tdx_event_irq", NULL)) { + pr_err("Event notification IRQ request failed\n"); + goto err_free_domain_irqs; + } + + cfg = irq_cfg(irq); + + if (_tdx_hypercall(TDVMCALL_SETUP_NOTIFY_INTR, cfg->vector, 0, 0, 0)) { + pr_err("Event notification hypercall failed\n"); + goto err_free_irqs; + } + + tdx_event_irq = irq; + + return 0; + +err_free_irqs: + free_irq(irq, NULL); +err_free_domain_irqs: + irq_domain_free_irqs(irq, 1); + + return -EIO; +} +early_initcall(tdx_event_irq_init) + +/** + * tdx_register_event_irq_cb() - Register TDX event IRQ callback handler. + * @handler: Address of driver specific event IRQ callback handler. Handler + * will be called in IRQ context and hence cannot sleep. + * @data: Context data to be passed to the callback handler. + * + * Return: 0 on success or standard error code on other failures. + */ +int tdx_register_event_irq_cb(tdx_event_irq_cb_t handler, void *data) +{ + struct event_irq_entry *entry; + unsigned long flags; + + if (tdx_event_irq <= 0) + return -EIO; + + entry = kzalloc(sizeof(*entry), GFP_KERNEL); + if (!entry) + return -ENOMEM; + + entry->data = data; + entry->handler = handler; + + spin_lock_irqsave(&event_irq_cb_lock, flags); + list_add_tail(&entry->head, &event_irq_cb_list); + spin_unlock_irqrestore(&event_irq_cb_lock, flags); + + return 0; +} +EXPORT_SYMBOL_GPL(tdx_register_event_irq_cb); + +/** + * tdx_unregister_event_irq_cb() - Unregister TDX event IRQ callback handler. + * @handler: Address of driver specific event IRQ callback handler. + * @data: Context data to be passed to the callback handler. + * + * Return: 0 on success or -EIO if event IRQ is not allocated. + */ +int tdx_unregister_event_irq_cb(tdx_event_irq_cb_t handler, void *data) +{ + struct event_irq_entry *entry; + unsigned long flags; + + if (tdx_event_irq <= 0) + return -EIO; + + spin_lock_irqsave(&event_irq_cb_lock, flags); + list_for_each_entry(entry, &event_irq_cb_list, head) { + if (entry->handler == handler && entry->data == data) { + list_del(&entry->head); + kfree(entry); + break; + } + } + spin_unlock_irqrestore(&event_irq_cb_lock, flags); + + return 0; +} +EXPORT_SYMBOL_GPL(tdx_unregister_event_irq_cb); diff --git a/arch/x86/include/asm/tdx.h b/arch/x86/include/asm/tdx.h index 28d889c9aa16..8807fe1b1f3f 100644 --- a/arch/x86/include/asm/tdx.h +++ b/arch/x86/include/asm/tdx.h @@ -53,6 +53,8 @@ struct ve_info {
#ifdef CONFIG_INTEL_TDX_GUEST
+typedef int (*tdx_event_irq_cb_t)(void *); + void __init tdx_early_init(void);
/* Used to communicate with the TDX module */ @@ -69,6 +71,10 @@ bool tdx_early_handle_ve(struct pt_regs *regs);
int tdx_mcall_get_report0(u8 *reportdata, u8 *tdreport);
+int tdx_register_event_irq_cb(tdx_event_irq_cb_t handler, void *data); + +int tdx_unregister_event_irq_cb(tdx_event_irq_cb_t handler, void *data); + #else
static inline void tdx_early_init(void) { };
On Sun, 2023-05-14 at 00:23 -0700, Kuppuswamy Sathyanarayanan wrote:
Host-guest event notification via configured interrupt vector is useful in cases where a guest makes an asynchronous request and needs a callback from the host to indicate the completion or to let the host notify the guest about events like device removal. One usage example is, callback requirement of GetQuote asynchronous hypercall.
In TDX guest, SetupEventNotifyInterrupt hypercall can be used by the guest to specify which interrupt vector to use as an event-notify vector from the VMM. Details about the SetupEventNotifyInterrupt hypercall can be found in TDX Guest-Host Communication Interface (GHCI) Specification, section "VP.VMCALL<SetupEventNotifyInterrupt>".
As per design, VMM will post the event completion IRQ using the same CPU on which SetupEventNotifyInterrupt hypercall request is received. So allocate an IRQ vector from "x86_vector_domain", and set the CPU affinity of the IRQ vector to the CPU on which SetupEventNotifyInterrupt hypercall is made.
Add tdx_register_event_irq_cb()/tdx_unregister_event_irq_cb() interfaces to allow drivers to register/unregister event notification handlers.
Reviewed-by: Tony Luck tony.luck@intel.com Reviewed-by: Andi Kleen ak@linux.intel.com Reviewed-by: Mika Westerberg mika.westerberg@linux.intel.com Reviewed-by: Erdem Aktas erdemaktas@google.com Acked-by: Kirill A. Shutemov kirill.shutemov@linux.intel.com Acked-by: Wander Lairson Costa wander@redhat.com Signed-off-by: Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com
Acked-by: Kai Huang kai.huang@intel.com
On Sun, May 14 2023 at 00:23, Kuppuswamy Sathyanarayanan wrote:
+static int __init tdx_event_irq_init(void) +{
- struct irq_affinity_desc desc;
- struct irq_alloc_info info;
- struct irq_cfg *cfg;
- int irq;
- if (!cpu_feature_enabled(X86_FEATURE_TDX_GUEST))
return 0;- init_irq_alloc_info(&info, NULL);
- cpumask_set_cpu(smp_processor_id(), &desc.mask);
desc is completely uninitialized here and therefore desc.mask and desc.is_managed contain random stack content.
The only reason why this "works" is that tdx_event_irq_init() is an early initcall which is invoked before SMP init, so the only choice of the allocator is to pick an interrupt on CPU0.
Thanks,
tglx
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest. The First step in the attestation process is TDREPORT generation, which involves getting the guest measurement data in the format of TDREPORT, which is further used to validate the authenticity of the TDX guest. TDREPORT by design is integrity-protected and can only be verified on the local machine.
To support remote verification of the TDREPORT (in a SGX-based attestation), the TDREPORT needs to be sent to the SGX Quoting Enclave (QE) to convert it to a remote verifiable Quote. SGX QE by design can only run outside of the TDX guest (i.e. in a host process or in a normal VM) and guest can use communication channels like vsock or TCP/IP to send the TDREPORT to the QE. But for security concerns, the TDX guest may not support these communication channels. To handle such cases, TDX defines a GetQuote hypercall which can be used by the guest to request the host VMM to communicate with the SGX QE. More details about GetQuote hypercall can be found in TDX Guest-Host Communication Interface (GHCI) for Intel TDX 1.0, section titled "TDG.VP.VMCALL<GetQuote>".
Add support for TDX_CMD_GET_QUOTE IOCTL to allow an attestation agent to submit GetQuote requests from the user space using GetQuote hypercall.
Since GetQuote is an asynchronous request hypercall, VMM will use the callback interrupt vector configured by the SetupEventNotifyInterrupt hypercall to notify the guest about Quote generation completion or failure. So register an IRQ handler for it.
GetQuote TDVMCALL requires TD guest pass a 4K aligned shared buffer with TDREPORT data as input, which is further used by the VMM to copy the TD Quote result after successful Quote generation. To create the shared buffer, allocate a large enough memory and mark it shared using set_memory_decrypted() in tdx_guest_init(). This buffer will be re-used for GetQuote requests in TDX_CMD_GET_QUOTE IOCTL handler.
Although this method reserves a fixed chunk of memory for GetQuote requests, such one-time allocation is preferable to the alternative choice of repeatedly allocating/freeing the shared buffer in the TDX_CMD_GET_QUOTE IOCTL handler, which will damage the direct map (because the sharing/unsharing process modifies the direct map). This allocation model is similar to that used by the AMD SEV guest driver.
Since the Quote generation process is not time-critical or frequently used, the current version does not support parallel GetQuote requests.
Reviewed-by: Tony Luck tony.luck@intel.com Reviewed-by: Andi Kleen ak@linux.intel.com Reviewed-by: Mika Westerberg mika.westerberg@linux.intel.com Reviewed-by: Erdem Aktas erdemaktas@google.com Acked-by: Kirill A. Shutemov kirill.shutemov@linux.intel.com Signed-off-by: Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com ---
Changes since v2: * Fixed commit log and comments as per review suggestion. * Renamed struct tdx_quote_hdr -> struct tdx_quote_buf. * Used alloc_pages_exact() instead of alloc_pages() in alloc_shared_pages(). * Clear quote buffer before each re-use in tdx_get_quote().
Changes since v1: * Removed platform bus device support. * Instead of allocating the shared buffers using DMA APIs in IOCTL handler, allocated it once in tdx_guest_init() and re-used it in GetQuote IOCTL handler. * To simplify the design, removed the support for parallel GetQuote requests. It can be added when there is a real requirement for it. * Fixed commit log and comments to reflect the latest changes.
Documentation/virt/coco/tdx-guest.rst | 11 ++ arch/x86/coco/tdx/tdx.c | 35 +++++ arch/x86/include/asm/tdx.h | 2 + drivers/virt/coco/tdx-guest/tdx-guest.c | 175 +++++++++++++++++++++++- include/uapi/linux/tdx-guest.h | 44 ++++++ 5 files changed, 266 insertions(+), 1 deletion(-)
diff --git a/Documentation/virt/coco/tdx-guest.rst b/Documentation/virt/coco/tdx-guest.rst index 46e316db6bb4..35e829466265 100644 --- a/Documentation/virt/coco/tdx-guest.rst +++ b/Documentation/virt/coco/tdx-guest.rst @@ -42,6 +42,17 @@ ABI. However, in the future, if the TDX Module supports more than one subtype, a new IOCTL CMD will be created to handle it. To keep the IOCTL naming consistent, a subtype index is added as part of the IOCTL CMD.
+2.2 TDX_CMD_GET_QUOTE +---------------------- + +:Input parameters: struct tdx_quote_req +:Output: Return 0 on success, -EIO on TDCALL failure or standard error number + on common failures. Upon successful execution, Quote data is copied + to tdx_quote_req.buf. + +The TDX_CMD_GET_QUOTE IOCTL can be used by attestation software to generate +Quote for the given TDREPORT using TDG.VP.VMCALL<GetQuote> hypercall. + Reference ---------
diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c index 49d7e11066c1..7e0af10515a6 100644 --- a/arch/x86/coco/tdx/tdx.c +++ b/arch/x86/coco/tdx/tdx.c @@ -33,6 +33,7 @@ #define TDVMCALL_MAP_GPA 0x10001 #define TDVMCALL_REPORT_FATAL_ERROR 0x10003 #define TDVMCALL_SETUP_NOTIFY_INTR 0x10004 +#define TDVMCALL_GET_QUOTE 0x10002
/* MMIO direction */ #define EPT_READ 0 @@ -198,6 +199,40 @@ static void __noreturn tdx_panic(const char *msg) __tdx_hypercall(&args); }
+/** + * tdx_hcall_get_quote() - Wrapper to request TD Quote using GetQuote + * hypercall. + * @buf: Address of the directly mapped shared kernel buffer which + * contains TDREPORT data. The same buffer will be used by + * VMM to store the generated TD Quote output. + * @size: size of the tdquote buffer (4KB-aligned). + * + * Refer to section titled "TDG.VP.VMCALL<GetQuote>" in the TDX GHCI + * v1.0 specification for more information on GetQuote hypercall. + * It is used in the TDX guest driver module to get the TD Quote. + * + * Return 0 on success or error code on failure. + */ +int tdx_hcall_get_quote(u8 *buf, size_t size) +{ + struct tdx_hypercall_args args = {0}; + + args.r10 = TDX_HYPERCALL_STANDARD; + args.r11 = TDVMCALL_GET_QUOTE; + /* Since buf is a shared memory, set the shared (decrypted) bits */ + args.r12 = cc_mkdec(virt_to_phys(buf)); + args.r13 = size; + + /* + * Pass the physical address of TDREPORT to the VMM and + * trigger the Quote generation. It is not a blocking + * call, hence completion of this request will be notified to + * the TD guest via a callback interrupt. + */ + return __tdx_hypercall(&args); +} +EXPORT_SYMBOL_GPL(tdx_hcall_get_quote); + static void tdx_parse_tdinfo(u64 *cc_mask) { struct tdx_module_output out; diff --git a/arch/x86/include/asm/tdx.h b/arch/x86/include/asm/tdx.h index 8807fe1b1f3f..254468448a1b 100644 --- a/arch/x86/include/asm/tdx.h +++ b/arch/x86/include/asm/tdx.h @@ -75,6 +75,8 @@ int tdx_register_event_irq_cb(tdx_event_irq_cb_t handler, void *data);
int tdx_unregister_event_irq_cb(tdx_event_irq_cb_t handler, void *data);
+int tdx_hcall_get_quote(u8 *buf, size_t size); + #else
static inline void tdx_early_init(void) { }; diff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 5e44a0fa69bd..388491fa63a1 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -12,12 +12,106 @@ #include <linux/mod_devicetable.h> #include <linux/string.h> #include <linux/uaccess.h> +#include <linux/set_memory.h>
#include <uapi/linux/tdx-guest.h>
#include <asm/cpu_device_id.h> #include <asm/tdx.h>
+/* + * Intel's SGX QE implementation generally uses Quote size less + * than 8K; Use 16K as MAX size to handle future updates and other + * 3rd party implementations. + */ +#define GET_QUOTE_MAX_SIZE (4 * PAGE_SIZE) + +/** + * struct quote_entry - Quote request struct + * @valid: Flag to check validity of the GetQuote request. + * @buf: Kernel buffer to share data with VMM (size is page aligned). + * @buf_len: Size of the buf in bytes. + * @compl: Completion object to track completion of GetQuote request. + */ +struct quote_entry { + bool valid; + void *buf; + size_t buf_len; + struct completion compl; +}; + +/* Quote data entry */ +static struct quote_entry *qentry; + +/* Lock to streamline quote requests */ +static DEFINE_MUTEX(quote_lock); + +static int quote_cb_handler(void *dev_id) +{ + struct quote_entry *entry = dev_id; + struct tdx_quote_buf *quote_buf = entry->buf; + + if (entry->valid && quote_buf->status != GET_QUOTE_IN_FLIGHT) + complete(&entry->compl); + + return 0; +} + +static void free_shared_pages(void *buf, size_t len) +{ + unsigned int count = PAGE_ALIGN(len) >> PAGE_SHIFT; + + set_memory_encrypted((unsigned long)buf, count); + + free_pages_exact(buf, PAGE_ALIGN(len)); +} + +static void *alloc_shared_pages(size_t len) +{ + unsigned int count = PAGE_ALIGN(len) >> PAGE_SHIFT; + void *addr; + int ret; + + addr = alloc_pages_exact(len, GFP_KERNEL); + if (!addr) + return NULL; + + ret = set_memory_decrypted((unsigned long)addr, count); + if (ret) { + free_pages_exact(addr, PAGE_ALIGN(len)); + return NULL; + } + + return addr; +} + +static struct quote_entry *alloc_quote_entry(size_t len) +{ + struct quote_entry *entry = NULL; + + entry = kmalloc(sizeof(*entry), GFP_KERNEL); + if (!entry) + return NULL; + + entry->buf = alloc_shared_pages(len); + if (!entry->buf) { + kfree(entry); + return NULL; + } + + entry->buf_len = PAGE_ALIGN(len); + init_completion(&entry->compl); + entry->valid = false; + + return entry; +} + +static void free_quote_entry(struct quote_entry *entry) +{ + free_shared_pages(entry->buf, entry->buf_len); + kfree(entry); +} + static long tdx_get_report0(struct tdx_report_req __user *req) { u8 *reportdata, *tdreport; @@ -53,12 +147,65 @@ static long tdx_get_report0(struct tdx_report_req __user *req) return ret; }
+static long tdx_get_quote(struct tdx_quote_req __user *ureq) +{ + struct tdx_quote_req req; + long ret; + + if (copy_from_user(&req, ureq, sizeof(req))) + return -EFAULT; + + mutex_lock("e_lock); + + if (!req.len || req.len > qentry->buf_len) { + ret = -EINVAL; + goto quote_failed; + } + + memset(qentry->buf, 0, qentry->buf_len); + reinit_completion(&qentry->compl); + qentry->valid = true; + + if (copy_from_user(qentry->buf, (void __user *)req.buf, req.len)) { + ret = -EFAULT; + goto quote_failed; + } + + /* Submit GetQuote Request using GetQuote hypercall */ + ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len); + if (ret) { + pr_err("GetQuote hypercall failed, status:%lx\n", ret); + ret = -EIO; + goto quote_failed; + } + + /* + * Although the GHCI specification does not state explicitly that + * the VMM must not wait indefinitely for the Quote request to be + * completed, a sane VMM should always notify the guest after a + * certain time, regardless of whether the Quote generation is + * successful or not. For now just assume the VMM will do so. + */ + wait_for_completion(&qentry->compl); + + if (copy_to_user((void __user *)req.buf, qentry->buf, req.len)) + ret = -EFAULT; + +quote_failed: + qentry->valid = false; + mutex_unlock("e_lock); + + return ret; +} + static long tdx_guest_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { switch (cmd) { case TDX_CMD_GET_REPORT0: return tdx_get_report0((struct tdx_report_req __user *)arg); + case TDX_CMD_GET_QUOTE: + return tdx_get_quote((struct tdx_quote_req *)arg); default: return -ENOTTY; } @@ -84,15 +231,41 @@ MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids);
static int __init tdx_guest_init(void) { + int ret; + if (!x86_match_cpu(tdx_guest_ids)) return -ENODEV;
- return misc_register(&tdx_misc_dev); + ret = misc_register(&tdx_misc_dev); + if (ret) + return ret; + + qentry = alloc_quote_entry(GET_QUOTE_MAX_SIZE); + if (!qentry) { + pr_err("Failed to allocate Quote buffer\n"); + ret = -ENOMEM; + goto free_misc; + } + + ret = tdx_register_event_irq_cb(quote_cb_handler, qentry); + if (ret) + goto free_quote; + + return 0; + +free_quote: + free_quote_entry(qentry); +free_misc: + misc_deregister(&tdx_misc_dev); + + return ret; } module_init(tdx_guest_init);
static void __exit tdx_guest_exit(void) { + tdx_unregister_event_irq_cb(quote_cb_handler, qentry); + free_quote_entry(qentry); misc_deregister(&tdx_misc_dev); } module_exit(tdx_guest_exit); diff --git a/include/uapi/linux/tdx-guest.h b/include/uapi/linux/tdx-guest.h index a6a2098c08ff..8a6ade299090 100644 --- a/include/uapi/linux/tdx-guest.h +++ b/include/uapi/linux/tdx-guest.h @@ -17,6 +17,12 @@ /* Length of TDREPORT used in TDG.MR.REPORT TDCALL */ #define TDX_REPORT_LEN 1024
+/* TD Quote status codes */ +#define GET_QUOTE_SUCCESS 0 +#define GET_QUOTE_IN_FLIGHT 0xffffffffffffffff +#define GET_QUOTE_ERROR 0x8000000000000000 +#define GET_QUOTE_SERVICE_UNAVAILABLE 0x8000000000000001 + /** * struct tdx_report_req - Request struct for TDX_CMD_GET_REPORT0 IOCTL. * @@ -30,6 +36,36 @@ struct tdx_report_req { __u8 tdreport[TDX_REPORT_LEN]; };
+/* struct tdx_quote_buf: Format of Quote request buffer. + * @version: Quote format version, filled by TD. + * @status: Status code of Quote request, filled by VMM. + * @in_len: Length of TDREPORT, filled by TD. + * @out_len: Length of Quote data, filled by VMM. + * @data: Quote data on output or TDREPORT on input. + * + * More details of Quote request buffer can be found in TDX + * Guest-Host Communication Interface (GHCI) for Intel TDX 1.0, + * section titled "TDG.VP.VMCALL<GetQuote>" + */ +struct tdx_quote_buf { + __u64 version; + __u64 status; + __u32 in_len; + __u32 out_len; + __u64 data[]; +}; + +/* struct tdx_quote_req: Request struct for TDX_CMD_GET_QUOTE IOCTL. + * @buf: Address of user buffer in the format of struct tdx_quote_buf. + * Upon successful completion of IOCTL, output is copied back to + * the same buffer (in struct tdx_quote_buf.data). + * @len: Length of the Quote buffer. + */ +struct tdx_quote_req { + __u64 buf; + __u64 len; +}; + /* * TDX_CMD_GET_REPORT0 - Get TDREPORT0 (a.k.a. TDREPORT subtype 0) using * TDCALL[TDG.MR.REPORT] @@ -39,4 +75,12 @@ struct tdx_report_req { */ #define TDX_CMD_GET_REPORT0 _IOWR('T', 1, struct tdx_report_req)
+/* + * TDX_CMD_GET_QUOTE - Get TD Guest Quote from QE/QGS using GetQuote + * TDVMCALL. + * + * Returns 0 on success or standard errno on other failures. + */ +#define TDX_CMD_GET_QUOTE _IOWR('T', 2, struct tdx_quote_req) + #endif /* _UAPI_LINUX_TDX_GUEST_H_ */
On Sun, 2023-05-14 at 00:23 -0700, Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest. The First step in the attestation process is TDREPORT generation, which involves getting the guest measurement data in the format of TDREPORT, which is further used to validate the authenticity of the TDX guest. TDREPORT by design is integrity-protected and can only be verified on the local machine.
To support remote verification of the TDREPORT (in a SGX-based attestation), the TDREPORT needs to be sent to the SGX Quoting Enclave (QE) to convert it to a remote verifiable Quote. SGX QE by design can only run outside of the TDX guest (i.e. in a host process or in a normal VM) and guest can use communication channels like vsock or TCP/IP to send the TDREPORT to the QE. But for security concerns, the TDX guest may not support these communication channels. To handle such cases, TDX defines a GetQuote hypercall which can be used by the guest to request the host VMM to communicate with the SGX QE. More details about GetQuote hypercall can be found in TDX Guest-Host Communication Interface (GHCI) for Intel TDX 1.0, section titled "TDG.VP.VMCALL<GetQuote>".
Add support for TDX_CMD_GET_QUOTE IOCTL to allow an attestation agent to submit GetQuote requests from the user space using GetQuote hypercall.
Since GetQuote is an asynchronous request hypercall, VMM will use the callback interrupt vector configured by the SetupEventNotifyInterrupt hypercall to notify the guest about Quote generation completion or failure. So register an IRQ handler for it.
GetQuote TDVMCALL requires TD guest pass a 4K aligned shared buffer with TDREPORT data as input, which is further used by the VMM to copy the TD Quote result after successful Quote generation. To create the shared buffer, allocate a large enough memory and mark it shared using set_memory_decrypted() in tdx_guest_init(). This buffer will be re-used for GetQuote requests in TDX_CMD_GET_QUOTE IOCTL handler.
Although this method reserves a fixed chunk of memory for GetQuote requests, such one-time allocation is preferable to the alternative choice of repeatedly allocating/freeing the shared buffer in the TDX_CMD_GET_QUOTE IOCTL handler, which will damage the direct map (because the sharing/unsharing process modifies the direct map). This allocation model is similar to that used by the AMD SEV guest driver.
Since the Quote generation process is not time-critical or frequently used, the current version does not support parallel GetQuote requests.
Reviewed-by: Tony Luck tony.luck@intel.com Reviewed-by: Andi Kleen ak@linux.intel.com Reviewed-by: Mika Westerberg mika.westerberg@linux.intel.com Reviewed-by: Erdem Aktas erdemaktas@google.com Acked-by: Kirill A. Shutemov kirill.shutemov@linux.intel.com Signed-off-by: Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com
Acked-by: Kai Huang kai.huang@intel.com
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
* Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL. * Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request. * Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
Reviewed-by: Tony Luck tony.luck@intel.com Reviewed-by: Andi Kleen ak@linux.intel.com Reviewed-by: Shuah Khan skhan@linuxfoundation.org Reviewed-by: Mika Westerberg mika.westerberg@linux.intel.com Reviewed-by: Erdem Aktas erdemaktas@google.com Acked-by: Kirill A. Shutemov kirill.shutemov@linux.intel.com Signed-off-by: Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com ---
Changes since v2: * Adapted to struct tdx_quote_hdr -> struct tdx_quote_buf rename.
tools/testing/selftests/tdx/tdx_guest_test.c | 65 ++++++++++++++++++-- 1 file changed, 59 insertions(+), 6 deletions(-)
diff --git a/tools/testing/selftests/tdx/tdx_guest_test.c b/tools/testing/selftests/tdx/tdx_guest_test.c index 81d8cb88ea1a..0b4b4293b9cb 100644 --- a/tools/testing/selftests/tdx/tdx_guest_test.c +++ b/tools/testing/selftests/tdx/tdx_guest_test.c @@ -18,6 +18,7 @@ #define TDX_GUEST_DEVNAME "/dev/tdx_guest" #define HEX_DUMP_SIZE 8 #define DEBUG 0 +#define QUOTE_SIZE 8192
/** * struct tdreport_type - Type header of TDREPORT_STRUCT. @@ -128,21 +129,29 @@ static void print_array_hex(const char *title, const char *prefix_str, printf("\n"); }
+/* Helper function to get TDREPORT */ +long get_tdreport0(int devfd, struct tdx_report_req *req) +{ + int i; + + /* Generate sample report data */ + for (i = 0; i < TDX_REPORTDATA_LEN; i++) + req->reportdata[i] = i; + + return ioctl(devfd, TDX_CMD_GET_REPORT0, req); +} + TEST(verify_report) { struct tdx_report_req req; struct tdreport *tdreport; - int devfd, i; + int devfd;
devfd = open(TDX_GUEST_DEVNAME, O_RDWR | O_SYNC); ASSERT_LT(0, devfd);
- /* Generate sample report data */ - for (i = 0; i < TDX_REPORTDATA_LEN; i++) - req.reportdata[i] = i; - /* Get TDREPORT */ - ASSERT_EQ(0, ioctl(devfd, TDX_CMD_GET_REPORT0, &req)); + ASSERT_EQ(0, get_tdreport0(devfd, &req));
if (DEBUG) { print_array_hex("\n\t\tTDX report data\n", "", @@ -160,4 +169,48 @@ TEST(verify_report) ASSERT_EQ(0, close(devfd)); }
+TEST(verify_quote) +{ + struct tdx_quote_buf *quote_buf; + struct tdx_report_req rep_req; + struct tdx_quote_req req; + __u64 quote_buf_size; + int devfd; + + /* Open attestation device */ + devfd = open(TDX_GUEST_DEVNAME, O_RDWR | O_SYNC); + + ASSERT_LT(0, devfd); + + /* Add size for quote header */ + quote_buf_size = sizeof(*quote_buf) + QUOTE_SIZE; + + /* Allocate quote buffer */ + quote_buf = (struct tdx_quote_buf *)malloc(quote_buf_size); + ASSERT_NE(NULL, quote_buf); + + /* Initialize GetQuote header */ + quote_buf->version = 1; + quote_buf->status = GET_QUOTE_SUCCESS; + quote_buf->in_len = TDX_REPORT_LEN; + quote_buf->out_len = 0; + + /* Get TDREPORT data */ + ASSERT_EQ(0, get_tdreport0(devfd, &rep_req)); + + /* Fill GetQuote request */ + memcpy(quote_buf->data, rep_req.tdreport, TDX_REPORT_LEN); + req.buf = (__u64)quote_buf; + req.len = quote_buf_size; + + ASSERT_EQ(0, ioctl(devfd, TDX_CMD_GET_QUOTE, &req)); + + /* Check whether GetQuote request is successful */ + EXPECT_EQ(0, quote_buf->status); + + free(quote_buf); + + ASSERT_EQ(0, close(devfd)); +} + TEST_HARNESS_MAIN
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
My observation of SNP_GET_REPORT and TDX_CMD_GET_REPORT is that they are both passing blobs across the user/kernel and platform/kernel boundary for the purposes of unlocking other resources. To me that is a flow that the Keys subsystem has infrastructure to handle. It has the concept of upcalls and asynchronous population of blobs by handles and mechanisms to protect and cache those communications. Linux / the Keys subsystem could benefit from the enhancements it would need to cover these 2 cases. Specifically, the benefit that when ARM and RISC-V arrive with similar communications with platform TSMs (Trusted Security Module) they can build upon the same infrastructure.
David, am I reaching with that association? My strawman mapping of TDX_CMD_GET_QUOTE to request_key() is something like:
request_key(coco_quote, "description", "<uuencoded tdreport>")
Where this is a common key_type for all vendors, but the description and arguments have room for vendor differentiation when doing the upcall to the platform TSM, but userspace never needs to contend with the different vendor formats, that is all handled internally to the kernel.
At this point I am just looking for confirmation that the "every vendor invent a new character device + ioctl" does not scale and a deeper conversation is needed. Keys is a plausible solution to that ABI proliferation problem.
Hi Dan,
On 6/12/23 12:03 PM, Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
My observation of SNP_GET_REPORT and TDX_CMD_GET_REPORT is that they are both passing blobs across the user/kernel and platform/kernel boundary for the purposes of unlocking other resources. To me that is a flow that the Keys subsystem has infrastructure to handle. It has the concept of upcalls and asynchronous population of blobs by handles and mechanisms to protect and cache those communications. Linux / the Keys subsystem could benefit from the enhancements it would need to cover these 2 cases. Specifically, the benefit that when ARM and RISC-V arrive with similar communications with platform TSMs (Trusted Security Module) they can build upon the same infrastructure.
David, am I reaching with that association? My strawman mapping of TDX_CMD_GET_QUOTE to request_key() is something like:
request_key(coco_quote, "description", "<uuencoded tdreport>")
Where this is a common key_type for all vendors, but the description and arguments have room for vendor differentiation when doing the upcall to the platform TSM, but userspace never needs to contend with the different vendor formats, that is all handled internally to the kernel.
At this point I am just looking for confirmation that the "every vendor invent a new character device + ioctl" does not scale and a deeper conversation is needed. Keys is a plausible solution to that ABI proliferation problem.
I agree that vendor-specific interfaces do not scale, and the ABI generalization will benefit future vendors who require similar feature support. However, such generalization, in my opinion, will make more sense if the requirements at the top level are also generalized. Currently, each vendor has their own attestation flow, and the user ABI they introduced includes a lot of vendor-specific information to support it. IMO, it is difficult to hide these vendor-specific information from the user without generalizing the high level attestation flow.
I have included the attestation IOCTL interfaces used by S390, AMD SEV and TDX below for reference. As you can see, each of these ARCHs uses very different input and output data formats. It contains a lot of vendor-specific information (like the vmpl field in struct snp_report_req or the meas_addr, arcb_adr in struct uvio_attest). The only thing I see in common is the use of input and output blobs.
Even if we just generalize the ABI now, I'm not sure if the unified ABI we create now will meet the requirements of RISC-v or ARM platforms when they introduce their own attestation flow in the future. My thinking is that without some sort of arch-agnostic high-level attestation workflow, ABI generalization has very little benefit.
====================================================================== // Following are S390 specific attestation struct drivers/s390/char/uvdevice.c
struct uvio_ioctl_cb { __u32 flags; __u16 uv_rc; /* UV header rc value */ __u16 uv_rrc; /* UV header rrc value */ __u64 argument_addr; /* Userspace address of uvio argument */ __u32 argument_len; __u8 reserved14[0x40 - 0x14]; /* must be zero */ };
#define UVIO_ATT_USER_DATA_LEN 0x100 #define UVIO_ATT_UID_LEN 0x10
struct uvio_attest { __u64 arcb_addr; /* 0x0000 */ __u64 meas_addr; /* 0x0008 */ __u64 add_data_addr; /* 0x0010 */ __u8 user_data[UVIO_ATT_USER_DATA_LEN]; /* 0x0018 */ __u8 config_uid[UVIO_ATT_UID_LEN]; /* 0x0118 */ __u32 arcb_len; /* 0x0128 */ __u32 meas_len; /* 0x012c */ __u32 add_data_len; /* 0x0130 */ __u16 user_data_len; /* 0x0134 */ __u16 reserved136; /* 0x0136 */ };
#define UVIO_DEVICE_NAME "uv" #define UVIO_TYPE_UVC 'u'
#define UVIO_IOCTL_ATT _IOWR(UVIO_TYPE_UVC, 0x01, struct uvio_ioctl_cb)
// Following are TDX specific interfaces drivers/virt/coco/tdx-guest/tdx-guest.c
/** * struct tdx_report_req - Request struct for TDX_CMD_GET_REPORT0 IOCTL. * * @reportdata: User buffer with REPORTDATA to be included into TDREPORT. * Typically it can be some nonce provided by attestation * service, so the generated TDREPORT can be uniquely verified. * @tdreport: User buffer to store TDREPORT output from TDCALL[TDG.MR.REPORT]. */ struct tdx_report_req { __u8 reportdata[TDX_REPORTDATA_LEN]; __u8 tdreport[TDX_REPORT_LEN]; };
/* struct tdx_quote_buf: Format of Quote request buffer. * @version: Quote format version, filled by TD. * @status: Status code of Quote request, filled by VMM. * @in_len: Length of TDREPORT, filled by TD. * @out_len: Length of Quote data, filled by VMM. * @data: Quote data on output or TDREPORT on input. * * More details of Quote request buffer can be found in TDX * Guest-Host Communication Interface (GHCI) for Intel TDX 1.0, * section titled "TDG.VP.VMCALL<GetQuote>" */ struct tdx_quote_buf { __u64 version; __u64 status; __u32 in_len; __u32 out_len; __u64 data[]; };
/* struct tdx_quote_req: Request struct for TDX_CMD_GET_QUOTE IOCTL. * @buf: Address of user buffer in the format of struct tdx_quote_buf. * Upon successful completion of IOCTL, output is copied back to * the same buffer (in struct tdx_quote_buf.data). * @len: Length of the Quote buffer. */ struct tdx_quote_req { __u64 buf; __u64 len; };
/* * TDX_CMD_GET_REPORT0 - Get TDREPORT0 (a.k.a. TDREPORT subtype 0) using * TDCALL[TDG.MR.REPORT] * * Return 0 on success, -EIO on TDCALL execution failure, and * standard errno on other general error cases. */ #define TDX_CMD_GET_REPORT0 _IOWR('T', 1, struct tdx_report_req)
/* * TDX_CMD_GET_QUOTE - Get TD Guest Quote from QE/QGS using GetQuote * TDVMCALL. * * Returns 0 on success or standard errno on other failures. */ #define TDX_CMD_GET_QUOTE _IOWR('T', 2, struct tdx_quote_req)
// Following are AMD SEV specific interfaces drivers/virt/coco/sev-guest/sev-guest.c
struct snp_report_req { /* user data that should be included in the report */ __u8 user_data[64];
/* The vmpl level to be included in the report */ __u32 vmpl;
/* Must be zero filled */ __u8 rsvd[28]; };
struct snp_report_resp { /* response data, see SEV-SNP spec for the format */ __u8 data[4000]; };
struct snp_guest_request_ioctl { /* message version number (must be non-zero) */ __u8 msg_version;
/* Request and response structure address */ __u64 req_data; __u64 resp_data;
/* bits[63:32]: VMM error code, bits[31:0] firmware error code (see psp-sev.h) */ union { __u64 exitinfo2; struct { __u32 fw_error; __u32 vmm_error; }; }; };
/* Get SNP attestation report */ #define SNP_GET_REPORT _IOWR(SNP_GUEST_REQ_IOC_TYPE, 0x0, struct snp_guest_request_ioctl) ======================================================================
In addition to GetReport support, each of these guest attestation drivers includes IOCTLs to handle vendor-specific needs (such as Derived key support in the SEV driver or RTMR Extend support in the TDX guest driver). So we cannot completely unify all IOCTL interfaces in these drivers.
On Mon, Jun 12, 2023 at 12:03 PM Dan Williams dan.j.williams@intel.com wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple
Compared to the number of SCSI vendors, I think the number of CPU vendors for confidential computing seems manageable to me. Is this really a good comparison?
confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
<Just my personal opinion below> I agree with this statement in the high level but it is also somehow surprising for me after all the discussion happened around this topic. Honestly, I feel like there are multiple versions of "Intel" working in different directions.
If we want multiple vendors trying to do the similar things behind a common ABI, it should start with the spec. Since this comment is coming from Intel, I wonder if there is any plan to combine the GHCB and GHCI interfaces under common ABI in the future or why it did not even happen in the first place.
What I see is that Intel has GETQUOTE TDVMCALL interface in its spec and again Intel does not really want to provide support for it in linux. It feels really frustrating.
My observation of SNP_GET_REPORT and TDX_CMD_GET_REPORT is that they are both passing blobs across the user/kernel and platform/kernel boundary for the purposes of unlocking other resources. To me that is a flow that the Keys subsystem has infrastructure to handle. It has the concept of upcalls and asynchronous population of blobs by handles and mechanisms to protect and cache those communications. Linux / the Keys subsystem could benefit from the enhancements it would need to cover these 2 cases. Specifically, the benefit that when ARM and RISC-V arrive with similar communications with platform TSMs (Trusted Security Module) they can build upon the same infrastructure.
David, am I reaching with that association? My strawman mapping of TDX_CMD_GET_QUOTE to request_key() is something like:
request_key(coco_quote, "description", "<uuencoded tdreport>")
Where this is a common key_type for all vendors, but the description and arguments have room for vendor differentiation when doing the upcall to the platform TSM, but userspace never needs to contend with the different vendor formats, that is all handled internally to the kernel.
I think the problem definition here is not accurate. With AMD SNP, guests need to do a hypercall to KVM and KVM needs to issue a SNP_GUEST_REQUEST(MSG_REPORT_REQ) to the SP firmware. In TDX, guests need to do a TDCALL to TDXMODULE to get the TDREPORT and then it needs to get that report delivered to the host userspace to get the TDQUOTE generated by the SGX quoting enclave. Also TDQUOTE is designed to work async while the SNP_GUEST_REQUESTS are blocking vmcalls.
Those are completely different flows. Are you suggesting that intel should also come down to a single call to get the TDQUOTE like AMD SNP?
The TDCALL interface asking for the TDREPORT is already there. AMD does not need to ask the report and the quote separately.
Here, the problem was that Intel (or "upstream community") did not want to implement/accept hypercall for TDQUOTE which would be handled by the user space VMM. The alternative implementation (using vsock) does not work for many use cases including ours. I do not see how your suggestion addresses the problem that this patch was trying to solve.
So while I like the suggested direction, I am not sure how much it is possible to come up with a common ABI even with just only for 2 vendors (AMD and Intel) without doing spec changes which is a multi year effort imho.
At this point I am just looking for confirmation that the "every vendor invent a new character device + ioctl" does not scale and a deeper conversation is needed. Keys is a plausible solution to that ABI proliferation problem.
On Thu, 2023-06-22 at 16:31 -0700, Erdem Aktas wrote:
So while I like the suggested direction, I am not sure how much it is possible to come up with a common ABI even with just only for 2 vendors (AMD and Intel) without doing spec changes which is a multi year effort imho.
I don't want to intervene the discussion around whether this direction is correct or not, however I want to say request_key() may not be the right place to fit Quote (or remote verifiable data blob in general for attestation).
request_key(coco_quote, "description", "<uuencoded tdreport>")
Although both key and Quote are data blob in some way, Quote certainly is not a key but have much more information. The man page of request_key() seems to suggest it's just for key:
request_key - request a key from the kernel's key management facility
So IMHO using request_key() to fit Quote may cause bigger confusion.
Huang, Kai wrote:
On Thu, 2023-06-22 at 16:31 -0700, Erdem Aktas wrote:
So while I like the suggested direction, I am not sure how much it is possible to come up with a common ABI even with just only for 2 vendors (AMD and Intel) without doing spec changes which is a multi year effort imho.
I don't want to intervene the discussion around whether this direction is correct or not, however I want to say request_key() may not be the right place to fit Quote (or remote verifiable data blob in general for attestation).
request_key(coco_quote, "description", "<uuencoded tdreport>")
Although both key and Quote are data blob in some way, Quote certainly is not a key but have much more information. The man page of request_key() seems to suggest it's just for key:
request_key - request a key from the kernel's key management facility
Read further in that man page and see the example of generic user defined value stored as a "key". A "key" is just a blob that has meaning to access other resources / instantiate other keys.
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| * 00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
--- diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig"
source "drivers/virt/coco/tdx-guest/Kconfig"
+config GUEST_ATTEST + tristate + select KEYS + endif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h> + +#include <keys/request_key_auth-type.h> +#include <keys/user-type.h> + +#include "guest-attest.h" + +static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem); + +static struct guest_attest_ops *fetch_ops(void) +{ + return list_first_entry_or_null(&guest_attest_list, + struct guest_attest_ops, list); +} + +static struct guest_attest_ops *get_ops(void) +{ + down_read(&guest_attest_rwsem); + return fetch_ops(); +} + +static void put_ops(void) +{ + up_read(&guest_attest_rwsem); +} + +int register_guest_attest_ops(struct guest_attest_ops *ops) +{ + struct guest_attest_ops *conflict; + int rc; + + down_write(&guest_attest_rwsem); + conflict = fetch_ops(); + if (conflict) { + pr_err(""%s" ops already registered\n", conflict->name); + rc = -EEXIST; + goto out; + } + list_add(&ops->list, &guest_attest_list); + try_module_get(ops->module); + rc = 0; +out: + up_write(&guest_attest_rwsem); + return rc; +} +EXPORT_SYMBOL_GPL(register_guest_attest_ops); + +void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{ + down_write(&guest_attest_rwsem); + list_del(&ops->list); + up_write(&guest_attest_rwsem); + module_put(ops->module); +} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops); + +static int __guest_attest_request_key(struct key *key, int level, + struct key *dest_keyring, + const char *callout_info, int callout_len, + struct key *authkey) +{ + struct guest_attest_ops *ops; + void *payload = NULL; + int rc, payload_len; + + ops = get_ops(); + if (!ops) + return -ENOKEY; + + payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL); + if (!payload) { + rc = -ENOMEM; + goto out; + } + + payload_len = base64_decode(callout_info, callout_len, payload); + if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) { + rc = -EINVAL; + goto out; + } + + rc = ops->request_attest(key, level, dest_keyring, payload, payload_len, + authkey); +out: + kfree(payload); + put_ops(); + return rc; +} + +static int guest_attest_request_key(struct key *authkey, void *data) +{ + struct request_key_auth *rka = get_request_key_auth(authkey); + struct key *key = rka->target_key; + unsigned long long id; + int rc, level; + + pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op, + rka->callout_info ? (char *)rka->callout_info : ""none""); + + if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2) + return -EINVAL; + + if (!rka->callout_info) { + rc = -EINVAL; + goto out; + } + + rc = __guest_attest_request_key(key, level, rka->dest_keyring, + rka->callout_info, rka->callout_len, + authkey); +out: + complete_request_key(authkey, rc); + return rc; +} + +static int guest_attest_vet_description(const char *desc) +{ + unsigned long long id; + int level; + + if (sscanf(desc, "guest_attest:%d:%llu", &level, &id) != 2) + return -EINVAL; + return 0; +} + +static struct key_type key_type_guest_attest = { + .name = "guest_attest", + .preparse = user_preparse, + .free_preparse = user_free_preparse, + .instantiate = generic_key_instantiate, + .revoke = user_revoke, + .destroy = user_destroy, + .describe = user_describe, + .read = user_read, + .vet_description = guest_attest_vet_description, + .request_key = guest_attest_request_key, +}; + +static int __init guest_attest_init(void) +{ + return register_key_type(&key_type_guest_attest); +} + +static void __exit guest_attest_exit(void) +{ + unregister_key_type(&key_type_guest_attest); +} + +module_init(guest_attest_init); +module_exit(guest_attest_exit); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/virt/coco/tdx-guest/Kconfig b/drivers/virt/coco/tdx-guest/Kconfig index 14246fc2fb02..9a1ec85369fe 100644 --- a/drivers/virt/coco/tdx-guest/Kconfig +++ b/drivers/virt/coco/tdx-guest/Kconfig @@ -1,6 +1,7 @@ config TDX_GUEST_DRIVER tristate "TDX Guest driver" depends on INTEL_TDX_GUEST + select GUEST_ATTEST help The driver provides userspace interface to communicate with the TDX module to request the TDX guest details like attestation diff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 388491fa63a1..65b5aab284d9 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -13,11 +13,13 @@ #include <linux/string.h> #include <linux/uaccess.h> #include <linux/set_memory.h> +#include <linux/key-type.h>
#include <uapi/linux/tdx-guest.h>
#include <asm/cpu_device_id.h> #include <asm/tdx.h> +#include "../guest-attest/guest-attest.h"
/* * Intel's SGX QE implementation generally uses Quote size less @@ -229,6 +231,62 @@ static const struct x86_cpu_id tdx_guest_ids[] = { }; MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids);
+static int tdx_request_attest(struct key *key, int level, + struct key *dest_keyring, void *payload, + int payload_len, struct key *authkey) +{ + u8 *tdreport; + long ret; + + tdreport = kzalloc(TDX_REPORT_LEN, GFP_KERNEL); + if (!tdreport) + return -ENOMEM; + + /* Generate TDREPORT0 using "TDG.MR.REPORT" TDCALL */ + ret = tdx_mcall_get_report0(payload, tdreport); + if (ret) + goto out; + + mutex_lock("e_lock); + + memset(qentry->buf, 0, qentry->buf_len); + reinit_completion(&qentry->compl); + qentry->valid = true; + + /* Submit GetQuote Request using GetQuote hyperetall */ + ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len); + if (ret) { + pr_err("GetQuote hyperetall failed, status:%lx\n", ret); + ret = -EIO; + goto quote_failed; + } + + /* + * Although the GHCI specification does not state explicitly that + * the VMM must not wait indefinitely for the Quote request to be + * completed, a sane VMM should always notify the guest after a + * certain time, regardless of whether the Quote generation is + * successful or not. For now just assume the VMM will do so. + */ + wait_for_completion(&qentry->compl); + + ret = key_instantiate_and_link(key, qentry->buf, qentry->buf_len, + dest_keyring, authkey); + +quote_failed: + qentry->valid = false; + mutex_unlock("e_lock); +out: + kfree(tdreport); + return ret; +} + +static struct guest_attest_ops tdx_attest_ops = { + .name = KBUILD_MODNAME, + .module = THIS_MODULE, + .request_attest = tdx_request_attest, +}; + static int __init tdx_guest_init(void) { int ret; @@ -251,8 +309,14 @@ static int __init tdx_guest_init(void) if (ret) goto free_quote;
+ ret = register_guest_attest_ops(&tdx_attest_ops); + if (ret) + goto free_irq; + return 0;
+free_irq: + tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote: free_quote_entry(qentry); free_misc: @@ -264,6 +328,7 @@ module_init(tdx_guest_init);
static void __exit tdx_guest_exit(void) { + unregister_guest_attest_ops(&tdx_attest_ops); tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote_entry(qentry); misc_deregister(&tdx_misc_dev);
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the
IIUC, the idea here is to cache the quote data and return it to the user whenever possible, right? If yes, I think such optimization may not be very useful for our case. AFAIK, the quote data will change whenever there is a change in the guest measurement data. Since the validity of the generated quote will not be long, and the frequency of quote generation requests is expected to be less, we may not get much benefit from caching the quote data. I think we can keep this logic simple by directly retrieving the quote data from the quoting enclave whenever there is a request from the user.
top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
This is only useful for key-server use case, right? Attestation can also be used for use cases like pattern matching or uploading some secure data, etc. Since key-server is not the only use case, does it make sense to suppport this derived key feature?
diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig" source "drivers/virt/coco/tdx-guest/Kconfig" +config GUEST_ATTEST
- tristate
- select KEYS
endif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h>
+#include <keys/request_key_auth-type.h> +#include <keys/user-type.h>
+#include "guest-attest.h"
Can you share you guest-attest.h?
+static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem);
+static struct guest_attest_ops *fetch_ops(void) +{
- return list_first_entry_or_null(&guest_attest_list,
struct guest_attest_ops, list);+}
+static struct guest_attest_ops *get_ops(void) +{
- down_read(&guest_attest_rwsem);
- return fetch_ops();
+}
+static void put_ops(void) +{
- up_read(&guest_attest_rwsem);
+}
+int register_guest_attest_ops(struct guest_attest_ops *ops) +{
- struct guest_attest_ops *conflict;
- int rc;
- down_write(&guest_attest_rwsem);
- conflict = fetch_ops();
- if (conflict) {
pr_err("\"%s\" ops already registered\n", conflict->name);rc = -EEXIST;goto out;- }
- list_add(&ops->list, &guest_attest_list);
- try_module_get(ops->module);
- rc = 0;
+out:
- up_write(&guest_attest_rwsem);
- return rc;
+} +EXPORT_SYMBOL_GPL(register_guest_attest_ops);
+void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{
- down_write(&guest_attest_rwsem);
- list_del(&ops->list);
- up_write(&guest_attest_rwsem);
- module_put(ops->module);
+} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops);
+static int __guest_attest_request_key(struct key *key, int level,
struct key *dest_keyring,const char *callout_info, int callout_len,struct key *authkey)+{
- struct guest_attest_ops *ops;
- void *payload = NULL;
- int rc, payload_len;
- ops = get_ops();
- if (!ops)
return -ENOKEY;- payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL);
- if (!payload) {
rc = -ENOMEM;goto out;- }
Is the idea to get the values like vmpl part of the payload?
- payload_len = base64_decode(callout_info, callout_len, payload);
- if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) {
rc = -EINVAL;goto out;- }
- rc = ops->request_attest(key, level, dest_keyring, payload, payload_len,
authkey);+out:
- kfree(payload);
- put_ops();
- return rc;
+}
+static int guest_attest_request_key(struct key *authkey, void *data) +{
- struct request_key_auth *rka = get_request_key_auth(authkey);
- struct key *key = rka->target_key;
- unsigned long long id;
- int rc, level;
- pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op,
rka->callout_info ? (char *)rka->callout_info : "\"none\"");- if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2)
return -EINVAL;
Can you explain some details about the id and level? It is not very clear why we need it.
- if (!rka->callout_info) {
rc = -EINVAL;goto out;- }
- rc = __guest_attest_request_key(key, level, rka->dest_keyring,
rka->callout_info, rka->callout_len,authkey);+out:
- complete_request_key(authkey, rc);
- return rc;
+}
+static int guest_attest_vet_description(const char *desc) +{
- unsigned long long id;
- int level;
- if (sscanf(desc, "guest_attest:%d:%llu", &level, &id) != 2)
return -EINVAL;- return 0;
+}
+static struct key_type key_type_guest_attest = {
- .name = "guest_attest",
- .preparse = user_preparse,
- .free_preparse = user_free_preparse,
- .instantiate = generic_key_instantiate,
- .revoke = user_revoke,
- .destroy = user_destroy,
- .describe = user_describe,
- .read = user_read,
- .vet_description = guest_attest_vet_description,
- .request_key = guest_attest_request_key,
+};
+static int __init guest_attest_init(void) +{
- return register_key_type(&key_type_guest_attest);
+}
+static void __exit guest_attest_exit(void) +{
- unregister_key_type(&key_type_guest_attest);
+}
+module_init(guest_attest_init); +module_exit(guest_attest_exit); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/virt/coco/tdx-guest/Kconfig b/drivers/virt/coco/tdx-guest/Kconfig index 14246fc2fb02..9a1ec85369fe 100644 --- a/drivers/virt/coco/tdx-guest/Kconfig +++ b/drivers/virt/coco/tdx-guest/Kconfig @@ -1,6 +1,7 @@ config TDX_GUEST_DRIVER tristate "TDX Guest driver" depends on INTEL_TDX_GUEST
- select GUEST_ATTEST help The driver provides userspace interface to communicate with the TDX module to request the TDX guest details like attestation
diff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 388491fa63a1..65b5aab284d9 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -13,11 +13,13 @@ #include <linux/string.h> #include <linux/uaccess.h> #include <linux/set_memory.h> +#include <linux/key-type.h> #include <uapi/linux/tdx-guest.h> #include <asm/cpu_device_id.h> #include <asm/tdx.h> +#include "../guest-attest/guest-attest.h" /*
- Intel's SGX QE implementation generally uses Quote size less
@@ -229,6 +231,62 @@ static const struct x86_cpu_id tdx_guest_ids[] = { }; MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids); +static int tdx_request_attest(struct key *key, int level,
struct key *dest_keyring, void *payload,int payload_len, struct key *authkey)+{
- u8 *tdreport;
- long ret;
- tdreport = kzalloc(TDX_REPORT_LEN, GFP_KERNEL);
- if (!tdreport)
return -ENOMEM;- /* Generate TDREPORT0 using "TDG.MR.REPORT" TDCALL */
- ret = tdx_mcall_get_report0(payload, tdreport);
- if (ret)
goto out;- mutex_lock("e_lock);
- memset(qentry->buf, 0, qentry->buf_len);
- reinit_completion(&qentry->compl);
- qentry->valid = true;
- /* Submit GetQuote Request using GetQuote hyperetall */
- ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len);
- if (ret) {
pr_err("GetQuote hyperetall failed, status:%lx\n", ret);ret = -EIO;goto quote_failed;- }
- /*
* Although the GHCI specification does not state explicitly that* the VMM must not wait indefinitely for the Quote request to be* completed, a sane VMM should always notify the guest after a* certain time, regardless of whether the Quote generation is* successful or not. For now just assume the VMM will do so.*/- wait_for_completion(&qentry->compl);
- ret = key_instantiate_and_link(key, qentry->buf, qentry->buf_len,
dest_keyring, authkey);+quote_failed:
- qentry->valid = false;
- mutex_unlock("e_lock);
+out:
- kfree(tdreport);
- return ret;
+}
+static struct guest_attest_ops tdx_attest_ops = {
- .name = KBUILD_MODNAME,
- .module = THIS_MODULE,
- .request_attest = tdx_request_attest,
+};
static int __init tdx_guest_init(void) { int ret; @@ -251,8 +309,14 @@ static int __init tdx_guest_init(void) if (ret) goto free_quote;
- ret = register_guest_attest_ops(&tdx_attest_ops);
- if (ret)
goto free_irq;- return 0;
+free_irq:
- tdx_unregister_event_irq_cb(quote_cb_handler, qentry);
free_quote: free_quote_entry(qentry); free_misc: @@ -264,6 +328,7 @@ module_init(tdx_guest_init); static void __exit tdx_guest_exit(void) {
- unregister_guest_attest_ops(&tdx_attest_ops); tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote_entry(qentry); misc_deregister(&tdx_misc_dev);
On Sun, Jun 25, 2023 at 8:06 PM Sathyanarayanan Kuppuswamy sathyanarayanan.kuppuswamy@linux.intel.com wrote:
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology. I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
ARM's Confidential Computing Realm Management Extensions (RME) seems to be going along the lines of a runtime measurement register model with their hardware enforced security. The number of registers isn't prescribed in the spec.
+Joey Gouly +linux-coco@lists.linux.dev as far as RME is concerned, do you know who would be best to weigh in on this discussion of a unified attestation model?
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the
IIUC, the idea here is to cache the quote data and return it to the user whenever possible, right? If yes, I think such optimization may not be very useful for our case. AFAIK, the quote data will change whenever there is a change in the guest measurement data. Since the validity of the generated quote will not be long, and the frequency of quote generation requests is expected to be less, we may not get much benefit from caching the quote data. I think we can keep this logic simple by directly retrieving the quote data from the quoting enclave whenever there is a request from the user.
top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
This is only useful for key-server use case, right? Attestation can also be used for use cases like pattern matching or uploading some secure data, etc. Since key-server is not the only use case, does it make sense to suppport this derived key feature?
diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig"
source "drivers/virt/coco/tdx-guest/Kconfig"
+config GUEST_ATTEST
tristateselect KEYSendif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h>
+#include <keys/request_key_auth-type.h> +#include <keys/user-type.h>
+#include "guest-attest.h"
Can you share you guest-attest.h?
+static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem);
+static struct guest_attest_ops *fetch_ops(void) +{
return list_first_entry_or_null(&guest_attest_list,struct guest_attest_ops, list);+}
+static struct guest_attest_ops *get_ops(void) +{
down_read(&guest_attest_rwsem);return fetch_ops();+}
+static void put_ops(void) +{
up_read(&guest_attest_rwsem);+}
+int register_guest_attest_ops(struct guest_attest_ops *ops) +{
struct guest_attest_ops *conflict;int rc;down_write(&guest_attest_rwsem);conflict = fetch_ops();if (conflict) {pr_err("\"%s\" ops already registered\n", conflict->name);rc = -EEXIST;goto out;}list_add(&ops->list, &guest_attest_list);try_module_get(ops->module);rc = 0;+out:
up_write(&guest_attest_rwsem);return rc;+} +EXPORT_SYMBOL_GPL(register_guest_attest_ops);
+void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{
down_write(&guest_attest_rwsem);list_del(&ops->list);up_write(&guest_attest_rwsem);module_put(ops->module);+} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops);
+static int __guest_attest_request_key(struct key *key, int level,
struct key *dest_keyring,const char *callout_info, int callout_len,struct key *authkey)+{
struct guest_attest_ops *ops;void *payload = NULL;int rc, payload_len;ops = get_ops();if (!ops)return -ENOKEY;payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL);if (!payload) {rc = -ENOMEM;goto out;}Is the idea to get the values like vmpl part of the payload?
payload_len = base64_decode(callout_info, callout_len, payload);if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) {rc = -EINVAL;goto out;}rc = ops->request_attest(key, level, dest_keyring, payload, payload_len,authkey);+out:
kfree(payload);put_ops();return rc;+}
+static int guest_attest_request_key(struct key *authkey, void *data) +{
struct request_key_auth *rka = get_request_key_auth(authkey);struct key *key = rka->target_key;unsigned long long id;int rc, level;pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op,rka->callout_info ? (char *)rka->callout_info : "\"none\"");if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;Can you explain some details about the id and level? It is not very clear why we need it.
if (!rka->callout_info) {rc = -EINVAL;goto out;}rc = __guest_attest_request_key(key, level, rka->dest_keyring,rka->callout_info, rka->callout_len,authkey);+out:
complete_request_key(authkey, rc);return rc;+}
+static int guest_attest_vet_description(const char *desc) +{
unsigned long long id;int level;if (sscanf(desc, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;return 0;+}
+static struct key_type key_type_guest_attest = {
.name = "guest_attest",.preparse = user_preparse,.free_preparse = user_free_preparse,.instantiate = generic_key_instantiate,.revoke = user_revoke,.destroy = user_destroy,.describe = user_describe,.read = user_read,.vet_description = guest_attest_vet_description,.request_key = guest_attest_request_key,+};
+static int __init guest_attest_init(void) +{
return register_key_type(&key_type_guest_attest);+}
+static void __exit guest_attest_exit(void) +{
unregister_key_type(&key_type_guest_attest);+}
+module_init(guest_attest_init); +module_exit(guest_attest_exit); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/virt/coco/tdx-guest/Kconfig b/drivers/virt/coco/tdx-guest/Kconfig index 14246fc2fb02..9a1ec85369fe 100644 --- a/drivers/virt/coco/tdx-guest/Kconfig +++ b/drivers/virt/coco/tdx-guest/Kconfig @@ -1,6 +1,7 @@ config TDX_GUEST_DRIVER tristate "TDX Guest driver" depends on INTEL_TDX_GUEST
select GUEST_ATTEST help The driver provides userspace interface to communicate with the TDX module to request the TDX guest details like attestationdiff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 388491fa63a1..65b5aab284d9 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -13,11 +13,13 @@ #include <linux/string.h> #include <linux/uaccess.h> #include <linux/set_memory.h> +#include <linux/key-type.h>
#include <uapi/linux/tdx-guest.h>
#include <asm/cpu_device_id.h> #include <asm/tdx.h> +#include "../guest-attest/guest-attest.h"
/*
- Intel's SGX QE implementation generally uses Quote size less
@@ -229,6 +231,62 @@ static const struct x86_cpu_id tdx_guest_ids[] = { }; MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids);
+static int tdx_request_attest(struct key *key, int level,
struct key *dest_keyring, void *payload,int payload_len, struct key *authkey)+{
u8 *tdreport;long ret;tdreport = kzalloc(TDX_REPORT_LEN, GFP_KERNEL);if (!tdreport)return -ENOMEM;/* Generate TDREPORT0 using "TDG.MR.REPORT" TDCALL */ret = tdx_mcall_get_report0(payload, tdreport);if (ret)goto out;mutex_lock("e_lock);memset(qentry->buf, 0, qentry->buf_len);reinit_completion(&qentry->compl);qentry->valid = true;/* Submit GetQuote Request using GetQuote hyperetall */ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len);if (ret) {pr_err("GetQuote hyperetall failed, status:%lx\n", ret);ret = -EIO;goto quote_failed;}/** Although the GHCI specification does not state explicitly that* the VMM must not wait indefinitely for the Quote request to be* completed, a sane VMM should always notify the guest after a* certain time, regardless of whether the Quote generation is* successful or not. For now just assume the VMM will do so.*/wait_for_completion(&qentry->compl);ret = key_instantiate_and_link(key, qentry->buf, qentry->buf_len,dest_keyring, authkey);+quote_failed:
qentry->valid = false;mutex_unlock("e_lock);+out:
kfree(tdreport);return ret;+}
+static struct guest_attest_ops tdx_attest_ops = {
.name = KBUILD_MODNAME,.module = THIS_MODULE,.request_attest = tdx_request_attest,+};
static int __init tdx_guest_init(void) { int ret; @@ -251,8 +309,14 @@ static int __init tdx_guest_init(void) if (ret) goto free_quote;
ret = register_guest_attest_ops(&tdx_attest_ops);if (ret)goto free_irq;return 0;+free_irq:
tdx_unregister_event_irq_cb(quote_cb_handler, qentry);free_quote: free_quote_entry(qentry); free_misc: @@ -264,6 +328,7 @@ module_init(tdx_guest_init);
static void __exit tdx_guest_exit(void) {
unregister_guest_attest_ops(&tdx_attest_ops); tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote_entry(qentry); misc_deregister(&tdx_misc_dev);-- Sathyanarayanan Kuppuswamy Linux Kernel Developer
-- -Dionna Glaze, PhD (she/her)
+Atish
Atish, any comments on this topic from RISC-v?
On 6/26/23 11:57 AM, Dionna Amalie Glaze wrote:
On Sun, Jun 25, 2023 at 8:06 PM Sathyanarayanan Kuppuswamy sathyanarayanan.kuppuswamy@linux.intel.com wrote:
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology. I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
ARM's Confidential Computing Realm Management Extensions (RME) seems to be going along the lines of a runtime measurement register model with their hardware enforced security. The number of registers isn't prescribed in the spec.
+Joey Gouly +linux-coco@lists.linux.dev as far as RME is concerned, do you know who would be best to weigh in on this discussion of a unified attestation model?
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the
IIUC, the idea here is to cache the quote data and return it to the user whenever possible, right? If yes, I think such optimization may not be very useful for our case. AFAIK, the quote data will change whenever there is a change in the guest measurement data. Since the validity of the generated quote will not be long, and the frequency of quote generation requests is expected to be less, we may not get much benefit from caching the quote data. I think we can keep this logic simple by directly retrieving the quote data from the quoting enclave whenever there is a request from the user.
top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
This is only useful for key-server use case, right? Attestation can also be used for use cases like pattern matching or uploading some secure data, etc. Since key-server is not the only use case, does it make sense to suppport this derived key feature?
diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig"
source "drivers/virt/coco/tdx-guest/Kconfig"
+config GUEST_ATTEST
tristateselect KEYSendif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h>
+#include <keys/request_key_auth-type.h> +#include <keys/user-type.h>
+#include "guest-attest.h"
Can you share you guest-attest.h?
+static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem);
+static struct guest_attest_ops *fetch_ops(void) +{
return list_first_entry_or_null(&guest_attest_list,struct guest_attest_ops, list);+}
+static struct guest_attest_ops *get_ops(void) +{
down_read(&guest_attest_rwsem);return fetch_ops();+}
+static void put_ops(void) +{
up_read(&guest_attest_rwsem);+}
+int register_guest_attest_ops(struct guest_attest_ops *ops) +{
struct guest_attest_ops *conflict;int rc;down_write(&guest_attest_rwsem);conflict = fetch_ops();if (conflict) {pr_err("\"%s\" ops already registered\n", conflict->name);rc = -EEXIST;goto out;}list_add(&ops->list, &guest_attest_list);try_module_get(ops->module);rc = 0;+out:
up_write(&guest_attest_rwsem);return rc;+} +EXPORT_SYMBOL_GPL(register_guest_attest_ops);
+void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{
down_write(&guest_attest_rwsem);list_del(&ops->list);up_write(&guest_attest_rwsem);module_put(ops->module);+} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops);
+static int __guest_attest_request_key(struct key *key, int level,
struct key *dest_keyring,const char *callout_info, int callout_len,struct key *authkey)+{
struct guest_attest_ops *ops;void *payload = NULL;int rc, payload_len;ops = get_ops();if (!ops)return -ENOKEY;payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL);if (!payload) {rc = -ENOMEM;goto out;}Is the idea to get the values like vmpl part of the payload?
payload_len = base64_decode(callout_info, callout_len, payload);if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) {rc = -EINVAL;goto out;}rc = ops->request_attest(key, level, dest_keyring, payload, payload_len,authkey);+out:
kfree(payload);put_ops();return rc;+}
+static int guest_attest_request_key(struct key *authkey, void *data) +{
struct request_key_auth *rka = get_request_key_auth(authkey);struct key *key = rka->target_key;unsigned long long id;int rc, level;pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op,rka->callout_info ? (char *)rka->callout_info : "\"none\"");if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;Can you explain some details about the id and level? It is not very clear why we need it.
if (!rka->callout_info) {rc = -EINVAL;goto out;}rc = __guest_attest_request_key(key, level, rka->dest_keyring,rka->callout_info, rka->callout_len,authkey);+out:
complete_request_key(authkey, rc);return rc;+}
+static int guest_attest_vet_description(const char *desc) +{
unsigned long long id;int level;if (sscanf(desc, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;return 0;+}
+static struct key_type key_type_guest_attest = {
.name = "guest_attest",.preparse = user_preparse,.free_preparse = user_free_preparse,.instantiate = generic_key_instantiate,.revoke = user_revoke,.destroy = user_destroy,.describe = user_describe,.read = user_read,.vet_description = guest_attest_vet_description,.request_key = guest_attest_request_key,+};
+static int __init guest_attest_init(void) +{
return register_key_type(&key_type_guest_attest);+}
+static void __exit guest_attest_exit(void) +{
unregister_key_type(&key_type_guest_attest);+}
+module_init(guest_attest_init); +module_exit(guest_attest_exit); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/virt/coco/tdx-guest/Kconfig b/drivers/virt/coco/tdx-guest/Kconfig index 14246fc2fb02..9a1ec85369fe 100644 --- a/drivers/virt/coco/tdx-guest/Kconfig +++ b/drivers/virt/coco/tdx-guest/Kconfig @@ -1,6 +1,7 @@ config TDX_GUEST_DRIVER tristate "TDX Guest driver" depends on INTEL_TDX_GUEST
select GUEST_ATTEST help The driver provides userspace interface to communicate with the TDX module to request the TDX guest details like attestationdiff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 388491fa63a1..65b5aab284d9 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -13,11 +13,13 @@ #include <linux/string.h> #include <linux/uaccess.h> #include <linux/set_memory.h> +#include <linux/key-type.h>
#include <uapi/linux/tdx-guest.h>
#include <asm/cpu_device_id.h> #include <asm/tdx.h> +#include "../guest-attest/guest-attest.h"
/*
- Intel's SGX QE implementation generally uses Quote size less
@@ -229,6 +231,62 @@ static const struct x86_cpu_id tdx_guest_ids[] = { }; MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids);
+static int tdx_request_attest(struct key *key, int level,
struct key *dest_keyring, void *payload,int payload_len, struct key *authkey)+{
u8 *tdreport;long ret;tdreport = kzalloc(TDX_REPORT_LEN, GFP_KERNEL);if (!tdreport)return -ENOMEM;/* Generate TDREPORT0 using "TDG.MR.REPORT" TDCALL */ret = tdx_mcall_get_report0(payload, tdreport);if (ret)goto out;mutex_lock("e_lock);memset(qentry->buf, 0, qentry->buf_len);reinit_completion(&qentry->compl);qentry->valid = true;/* Submit GetQuote Request using GetQuote hyperetall */ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len);if (ret) {pr_err("GetQuote hyperetall failed, status:%lx\n", ret);ret = -EIO;goto quote_failed;}/** Although the GHCI specification does not state explicitly that* the VMM must not wait indefinitely for the Quote request to be* completed, a sane VMM should always notify the guest after a* certain time, regardless of whether the Quote generation is* successful or not. For now just assume the VMM will do so.*/wait_for_completion(&qentry->compl);ret = key_instantiate_and_link(key, qentry->buf, qentry->buf_len,dest_keyring, authkey);+quote_failed:
qentry->valid = false;mutex_unlock("e_lock);+out:
kfree(tdreport);return ret;+}
+static struct guest_attest_ops tdx_attest_ops = {
.name = KBUILD_MODNAME,.module = THIS_MODULE,.request_attest = tdx_request_attest,+};
static int __init tdx_guest_init(void) { int ret; @@ -251,8 +309,14 @@ static int __init tdx_guest_init(void) if (ret) goto free_quote;
ret = register_guest_attest_ops(&tdx_attest_ops);if (ret)goto free_irq;return 0;+free_irq:
tdx_unregister_event_irq_cb(quote_cb_handler, qentry);free_quote: free_quote_entry(qentry); free_misc: @@ -264,6 +328,7 @@ module_init(tdx_guest_init);
static void __exit tdx_guest_exit(void) {
unregister_guest_attest_ops(&tdx_attest_ops); tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote_entry(qentry); misc_deregister(&tdx_misc_dev);-- Sathyanarayanan Kuppuswamy Linux Kernel Developer
-- -Dionna Glaze, PhD (she/her)
On Mon, Jun 26, 2023 at 05:39:12PM -0700, Sathyanarayanan Kuppuswamy wrote:
+Atish
Atish, any comments on this topic from RISC-v?
The CoVE (RISC-V confidential computing specification) would benefit from the proposed abstract API. Similar to at least both TDX and SEV, the CoVE attestation evidence generation proposed (The spec is not ratified yet) interface [1] basically takes some binary blobs in (a TVM public key and an attestation challenge blob) and requests the TSM to generate an attesation evidence for the caller. The TSM will generate such evidence from all static and runtime measurements and sign it with its DICE derived key. Attestation lingo set aside, the pattern here is similar to SEV and TDX. Having a common API for a generic attestation evidence generation interface would avoid having to add yet another ioctl based interface specific to CoVE.
Another interface we could think about commonizing is the measurement extension one. I think both TDX and CoVE allow for a guest to dynamically extend its measurements (to dedicated, runtime PCRs).
Cheers, Samuel.
[1] https://github.com/riscv-non-isa/riscv-ap-tee/blob/main/specification/sbi_co...
On 6/26/23 11:57 AM, Dionna Amalie Glaze wrote:
On Sun, Jun 25, 2023 at 8:06 PM Sathyanarayanan Kuppuswamy sathyanarayanan.kuppuswamy@linux.intel.com wrote:
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology. I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
ARM's Confidential Computing Realm Management Extensions (RME) seems to be going along the lines of a runtime measurement register model with their hardware enforced security. The number of registers isn't prescribed in the spec.
+Joey Gouly +linux-coco@lists.linux.dev as far as RME is concerned, do you know who would be best to weigh in on this discussion of a unified attestation model?
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the
IIUC, the idea here is to cache the quote data and return it to the user whenever possible, right? If yes, I think such optimization may not be very useful for our case. AFAIK, the quote data will change whenever there is a change in the guest measurement data. Since the validity of the generated quote will not be long, and the frequency of quote generation requests is expected to be less, we may not get much benefit from caching the quote data. I think we can keep this logic simple by directly retrieving the quote data from the quoting enclave whenever there is a request from the user.
top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
This is only useful for key-server use case, right? Attestation can also be used for use cases like pattern matching or uploading some secure data, etc. Since key-server is not the only use case, does it make sense to suppport this derived key feature?
diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig"
source "drivers/virt/coco/tdx-guest/Kconfig"
+config GUEST_ATTEST
tristateselect KEYSendif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h>
+#include <keys/request_key_auth-type.h> +#include <keys/user-type.h>
+#include "guest-attest.h"
Can you share you guest-attest.h?
+static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem);
+static struct guest_attest_ops *fetch_ops(void) +{
return list_first_entry_or_null(&guest_attest_list,struct guest_attest_ops, list);+}
+static struct guest_attest_ops *get_ops(void) +{
down_read(&guest_attest_rwsem);return fetch_ops();+}
+static void put_ops(void) +{
up_read(&guest_attest_rwsem);+}
+int register_guest_attest_ops(struct guest_attest_ops *ops) +{
struct guest_attest_ops *conflict;int rc;down_write(&guest_attest_rwsem);conflict = fetch_ops();if (conflict) {pr_err("\"%s\" ops already registered\n", conflict->name);rc = -EEXIST;goto out;}list_add(&ops->list, &guest_attest_list);try_module_get(ops->module);rc = 0;+out:
up_write(&guest_attest_rwsem);return rc;+} +EXPORT_SYMBOL_GPL(register_guest_attest_ops);
+void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{
down_write(&guest_attest_rwsem);list_del(&ops->list);up_write(&guest_attest_rwsem);module_put(ops->module);+} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops);
+static int __guest_attest_request_key(struct key *key, int level,
struct key *dest_keyring,const char *callout_info, int callout_len,struct key *authkey)+{
struct guest_attest_ops *ops;void *payload = NULL;int rc, payload_len;ops = get_ops();if (!ops)return -ENOKEY;payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL);if (!payload) {rc = -ENOMEM;goto out;}Is the idea to get the values like vmpl part of the payload?
payload_len = base64_decode(callout_info, callout_len, payload);if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) {rc = -EINVAL;goto out;}rc = ops->request_attest(key, level, dest_keyring, payload, payload_len,authkey);+out:
kfree(payload);put_ops();return rc;+}
+static int guest_attest_request_key(struct key *authkey, void *data) +{
struct request_key_auth *rka = get_request_key_auth(authkey);struct key *key = rka->target_key;unsigned long long id;int rc, level;pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op,rka->callout_info ? (char *)rka->callout_info : "\"none\"");if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;Can you explain some details about the id and level? It is not very clear why we need it.
if (!rka->callout_info) {rc = -EINVAL;goto out;}rc = __guest_attest_request_key(key, level, rka->dest_keyring,rka->callout_info, rka->callout_len,authkey);+out:
complete_request_key(authkey, rc);return rc;+}
+static int guest_attest_vet_description(const char *desc) +{
unsigned long long id;int level;if (sscanf(desc, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;return 0;+}
+static struct key_type key_type_guest_attest = {
.name = "guest_attest",.preparse = user_preparse,.free_preparse = user_free_preparse,.instantiate = generic_key_instantiate,.revoke = user_revoke,.destroy = user_destroy,.describe = user_describe,.read = user_read,.vet_description = guest_attest_vet_description,.request_key = guest_attest_request_key,+};
+static int __init guest_attest_init(void) +{
return register_key_type(&key_type_guest_attest);+}
+static void __exit guest_attest_exit(void) +{
unregister_key_type(&key_type_guest_attest);+}
+module_init(guest_attest_init); +module_exit(guest_attest_exit); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/virt/coco/tdx-guest/Kconfig b/drivers/virt/coco/tdx-guest/Kconfig index 14246fc2fb02..9a1ec85369fe 100644 --- a/drivers/virt/coco/tdx-guest/Kconfig +++ b/drivers/virt/coco/tdx-guest/Kconfig @@ -1,6 +1,7 @@ config TDX_GUEST_DRIVER tristate "TDX Guest driver" depends on INTEL_TDX_GUEST
select GUEST_ATTEST help The driver provides userspace interface to communicate with the TDX module to request the TDX guest details like attestationdiff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 388491fa63a1..65b5aab284d9 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -13,11 +13,13 @@ #include <linux/string.h> #include <linux/uaccess.h> #include <linux/set_memory.h> +#include <linux/key-type.h>
#include <uapi/linux/tdx-guest.h>
#include <asm/cpu_device_id.h> #include <asm/tdx.h> +#include "../guest-attest/guest-attest.h"
/*
- Intel's SGX QE implementation generally uses Quote size less
@@ -229,6 +231,62 @@ static const struct x86_cpu_id tdx_guest_ids[] = { }; MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids);
+static int tdx_request_attest(struct key *key, int level,
struct key *dest_keyring, void *payload,int payload_len, struct key *authkey)+{
u8 *tdreport;long ret;tdreport = kzalloc(TDX_REPORT_LEN, GFP_KERNEL);if (!tdreport)return -ENOMEM;/* Generate TDREPORT0 using "TDG.MR.REPORT" TDCALL */ret = tdx_mcall_get_report0(payload, tdreport);if (ret)goto out;mutex_lock("e_lock);memset(qentry->buf, 0, qentry->buf_len);reinit_completion(&qentry->compl);qentry->valid = true;/* Submit GetQuote Request using GetQuote hyperetall */ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len);if (ret) {pr_err("GetQuote hyperetall failed, status:%lx\n", ret);ret = -EIO;goto quote_failed;}/** Although the GHCI specification does not state explicitly that* the VMM must not wait indefinitely for the Quote request to be* completed, a sane VMM should always notify the guest after a* certain time, regardless of whether the Quote generation is* successful or not. For now just assume the VMM will do so.*/wait_for_completion(&qentry->compl);ret = key_instantiate_and_link(key, qentry->buf, qentry->buf_len,dest_keyring, authkey);+quote_failed:
qentry->valid = false;mutex_unlock("e_lock);+out:
kfree(tdreport);return ret;+}
+static struct guest_attest_ops tdx_attest_ops = {
.name = KBUILD_MODNAME,.module = THIS_MODULE,.request_attest = tdx_request_attest,+};
static int __init tdx_guest_init(void) { int ret; @@ -251,8 +309,14 @@ static int __init tdx_guest_init(void) if (ret) goto free_quote;
ret = register_guest_attest_ops(&tdx_attest_ops);if (ret)goto free_irq;return 0;+free_irq:
tdx_unregister_event_irq_cb(quote_cb_handler, qentry);free_quote: free_quote_entry(qentry); free_misc: @@ -264,6 +328,7 @@ module_init(tdx_guest_init);
static void __exit tdx_guest_exit(void) {
unregister_guest_attest_ops(&tdx_attest_ops); tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote_entry(qentry); misc_deregister(&tdx_misc_dev);-- Sathyanarayanan Kuppuswamy Linux Kernel Developer
-- -Dionna Glaze, PhD (she/her)
-- Sathyanarayanan Kuppuswamy Linux Kernel Developer
Hi Samuel,
On 6/28/23 8:41 AM, Samuel Ortiz wrote:
On Mon, Jun 26, 2023 at 05:39:12PM -0700, Sathyanarayanan Kuppuswamy wrote:
+Atish
Atish, any comments on this topic from RISC-v?
The CoVE (RISC-V confidential computing specification) would benefit from the proposed abstract API. Similar to at least both TDX and SEV, the CoVE attestation evidence generation proposed (The spec is not ratified yet) interface [1] basically takes some binary blobs in (a TVM public key and an attestation challenge blob) and requests the TSM to generate an attesation evidence for the caller. The TSM will generate such evidence from all static and runtime measurements and sign it with its DICE derived key. Attestation lingo set aside, the pattern here is similar to SEV and TDX. Having a common API for a generic attestation evidence generation interface would avoid having to add yet another ioctl based interface specific to CoVE.
Great, this gives us more confidence about the generic ABI design.
Another interface we could think about commonizing is the measurement extension one. I think both TDX and CoVE allow for a guest to dynamically extend its measurements (to dedicated, runtime PCRs).
Yes. I think most vendors will need similar support. We are planning to add a generic ABI for this as well.
Cheers, Samuel.
[1] https://github.com/riscv-non-isa/riscv-ap-tee/blob/main/specification/sbi_co...
On 6/26/23 11:57 AM, Dionna Amalie Glaze wrote:
On Sun, Jun 25, 2023 at 8:06 PM Sathyanarayanan Kuppuswamy sathyanarayanan.kuppuswamy@linux.intel.com wrote:
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote: > In TDX guest, the second stage of the attestation process is Quote > generation. This process is required to convert the locally generated > TDREPORT into a remotely verifiable Quote. It involves sending the > TDREPORT data to a Quoting Enclave (QE) which will verify the > integrity of the TDREPORT and sign it with an attestation key. > > Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to > allow the user agent to get the TD Quote. > > Add a kernel selftest module to verify the Quote generation feature. > > TD Quote generation involves following steps: > > * Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL. > * Embed the TDREPORT data in quote buffer and request for quote > generation via TDX_CMD_GET_QUOTE IOCTL request. > * Upon completion of the GetQuote request, check for non zero value > in the status field of Quote header to make sure the generated > quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology. I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
ARM's Confidential Computing Realm Management Extensions (RME) seems to be going along the lines of a runtime measurement register model with their hardware enforced security. The number of registers isn't prescribed in the spec.
+Joey Gouly +linux-coco@lists.linux.dev as far as RME is concerned, do you know who would be best to weigh in on this discussion of a unified attestation model?
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the
IIUC, the idea here is to cache the quote data and return it to the user whenever possible, right? If yes, I think such optimization may not be very useful for our case. AFAIK, the quote data will change whenever there is a change in the guest measurement data. Since the validity of the generated quote will not be long, and the frequency of quote generation requests is expected to be less, we may not get much benefit from caching the quote data. I think we can keep this logic simple by directly retrieving the quote data from the quoting enclave whenever there is a request from the user.
top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
This is only useful for key-server use case, right? Attestation can also be used for use cases like pattern matching or uploading some secure data, etc. Since key-server is not the only use case, does it make sense to suppport this derived key feature?
diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig"
source "drivers/virt/coco/tdx-guest/Kconfig"
+config GUEST_ATTEST
tristateselect KEYSendif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h>
+#include <keys/request_key_auth-type.h> +#include <keys/user-type.h>
+#include "guest-attest.h"
Can you share you guest-attest.h?
+static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem);
+static struct guest_attest_ops *fetch_ops(void) +{
return list_first_entry_or_null(&guest_attest_list,struct guest_attest_ops, list);+}
+static struct guest_attest_ops *get_ops(void) +{
down_read(&guest_attest_rwsem);return fetch_ops();+}
+static void put_ops(void) +{
up_read(&guest_attest_rwsem);+}
+int register_guest_attest_ops(struct guest_attest_ops *ops) +{
struct guest_attest_ops *conflict;int rc;down_write(&guest_attest_rwsem);conflict = fetch_ops();if (conflict) {pr_err("\"%s\" ops already registered\n", conflict->name);rc = -EEXIST;goto out;}list_add(&ops->list, &guest_attest_list);try_module_get(ops->module);rc = 0;+out:
up_write(&guest_attest_rwsem);return rc;+} +EXPORT_SYMBOL_GPL(register_guest_attest_ops);
+void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{
down_write(&guest_attest_rwsem);list_del(&ops->list);up_write(&guest_attest_rwsem);module_put(ops->module);+} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops);
+static int __guest_attest_request_key(struct key *key, int level,
struct key *dest_keyring,const char *callout_info, int callout_len,struct key *authkey)+{
struct guest_attest_ops *ops;void *payload = NULL;int rc, payload_len;ops = get_ops();if (!ops)return -ENOKEY;payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL);if (!payload) {rc = -ENOMEM;goto out;}Is the idea to get the values like vmpl part of the payload?
payload_len = base64_decode(callout_info, callout_len, payload);if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) {rc = -EINVAL;goto out;}rc = ops->request_attest(key, level, dest_keyring, payload, payload_len,authkey);+out:
kfree(payload);put_ops();return rc;+}
+static int guest_attest_request_key(struct key *authkey, void *data) +{
struct request_key_auth *rka = get_request_key_auth(authkey);struct key *key = rka->target_key;unsigned long long id;int rc, level;pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op,rka->callout_info ? (char *)rka->callout_info : "\"none\"");if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;Can you explain some details about the id and level? It is not very clear why we need it.
if (!rka->callout_info) {rc = -EINVAL;goto out;}rc = __guest_attest_request_key(key, level, rka->dest_keyring,rka->callout_info, rka->callout_len,authkey);+out:
complete_request_key(authkey, rc);return rc;+}
+static int guest_attest_vet_description(const char *desc) +{
unsigned long long id;int level;if (sscanf(desc, "guest_attest:%d:%llu", &level, &id) != 2)return -EINVAL;return 0;+}
+static struct key_type key_type_guest_attest = {
.name = "guest_attest",.preparse = user_preparse,.free_preparse = user_free_preparse,.instantiate = generic_key_instantiate,.revoke = user_revoke,.destroy = user_destroy,.describe = user_describe,.read = user_read,.vet_description = guest_attest_vet_description,.request_key = guest_attest_request_key,+};
+static int __init guest_attest_init(void) +{
return register_key_type(&key_type_guest_attest);+}
+static void __exit guest_attest_exit(void) +{
unregister_key_type(&key_type_guest_attest);+}
+module_init(guest_attest_init); +module_exit(guest_attest_exit); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/virt/coco/tdx-guest/Kconfig b/drivers/virt/coco/tdx-guest/Kconfig index 14246fc2fb02..9a1ec85369fe 100644 --- a/drivers/virt/coco/tdx-guest/Kconfig +++ b/drivers/virt/coco/tdx-guest/Kconfig @@ -1,6 +1,7 @@ config TDX_GUEST_DRIVER tristate "TDX Guest driver" depends on INTEL_TDX_GUEST
select GUEST_ATTEST help The driver provides userspace interface to communicate with the TDX module to request the TDX guest details like attestationdiff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c index 388491fa63a1..65b5aab284d9 100644 --- a/drivers/virt/coco/tdx-guest/tdx-guest.c +++ b/drivers/virt/coco/tdx-guest/tdx-guest.c @@ -13,11 +13,13 @@ #include <linux/string.h> #include <linux/uaccess.h> #include <linux/set_memory.h> +#include <linux/key-type.h>
#include <uapi/linux/tdx-guest.h>
#include <asm/cpu_device_id.h> #include <asm/tdx.h> +#include "../guest-attest/guest-attest.h"
/*
- Intel's SGX QE implementation generally uses Quote size less
@@ -229,6 +231,62 @@ static const struct x86_cpu_id tdx_guest_ids[] = { }; MODULE_DEVICE_TABLE(x86cpu, tdx_guest_ids);
+static int tdx_request_attest(struct key *key, int level,
struct key *dest_keyring, void *payload,int payload_len, struct key *authkey)+{
u8 *tdreport;long ret;tdreport = kzalloc(TDX_REPORT_LEN, GFP_KERNEL);if (!tdreport)return -ENOMEM;/* Generate TDREPORT0 using "TDG.MR.REPORT" TDCALL */ret = tdx_mcall_get_report0(payload, tdreport);if (ret)goto out;mutex_lock("e_lock);memset(qentry->buf, 0, qentry->buf_len);reinit_completion(&qentry->compl);qentry->valid = true;/* Submit GetQuote Request using GetQuote hyperetall */ret = tdx_hcall_get_quote(qentry->buf, qentry->buf_len);if (ret) {pr_err("GetQuote hyperetall failed, status:%lx\n", ret);ret = -EIO;goto quote_failed;}/** Although the GHCI specification does not state explicitly that* the VMM must not wait indefinitely for the Quote request to be* completed, a sane VMM should always notify the guest after a* certain time, regardless of whether the Quote generation is* successful or not. For now just assume the VMM will do so.*/wait_for_completion(&qentry->compl);ret = key_instantiate_and_link(key, qentry->buf, qentry->buf_len,dest_keyring, authkey);+quote_failed:
qentry->valid = false;mutex_unlock("e_lock);+out:
kfree(tdreport);return ret;+}
+static struct guest_attest_ops tdx_attest_ops = {
.name = KBUILD_MODNAME,.module = THIS_MODULE,.request_attest = tdx_request_attest,+};
static int __init tdx_guest_init(void) { int ret; @@ -251,8 +309,14 @@ static int __init tdx_guest_init(void) if (ret) goto free_quote;
ret = register_guest_attest_ops(&tdx_attest_ops);if (ret)goto free_irq;return 0;+free_irq:
tdx_unregister_event_irq_cb(quote_cb_handler, qentry);free_quote: free_quote_entry(qentry); free_misc: @@ -264,6 +328,7 @@ module_init(tdx_guest_init);
static void __exit tdx_guest_exit(void) {
unregister_guest_attest_ops(&tdx_attest_ops); tdx_unregister_event_irq_cb(quote_cb_handler, qentry); free_quote_entry(qentry); misc_deregister(&tdx_misc_dev);-- Sathyanarayanan Kuppuswamy Linux Kernel Developer
-- -Dionna Glaze, PhD (she/her)
-- Sathyanarayanan Kuppuswamy Linux Kernel Developer
Dionna Amalie Glaze wrote:
On Sun, Jun 25, 2023 at 8:06 PM Sathyanarayanan Kuppuswamy sathyanarayanan.kuppuswamy@linux.intel.com wrote:
[..]
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology.
That's where I hope the line can be drawn, i.e. that all of this vendor differentiation really only matters inside the kernel in the end.
I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
This is where I need help. Can you identify where the following assertion falls over:
"The minimum viable key-server is one that can generically validate a blob with an ECDSA signature".
I.e. the fact that SEV and TDX send different length blobs is less important than validating that signature.
If it is always the case that specific fields in the blob need to be decoded then yes, that weakens the assertion. However, maybe that means that kernel code parses the blob and conveys that parsed info along with vendor attestation payload all signed by a Linux key. I.e. still allow for a unified output format + signed vendor blob and provide a path to keep all the vendor specific handling internal to the kernel.
ARM's Confidential Computing Realm Management Extensions (RME) seems to be going along the lines of a runtime measurement register model with their hardware enforced security. The number of registers isn't prescribed in the spec.
+Joey Gouly +linux-coco@lists.linux.dev as far as RME is concerned, do you know who would be best to weigh in on this discussion of a unified attestation model?
On Tue, Jun 27, 2023 at 5:13 PM Dan Williams dan.j.williams@intel.com wrote:
[..]
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology.
That's where I hope the line can be drawn, i.e. that all of this vendor differentiation really only matters inside the kernel in the end.
I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
This is where I need help. Can you identify where the following assertion falls over:
"The minimum viable key-server is one that can generically validate a blob with an ECDSA signature".
I.e. the fact that SEV and TDX send different length blobs is less important than validating that signature.
If it is always the case that specific fields in the blob need to be decoded then yes, that weakens the assertion. However, maybe that means that kernel code parses the blob and conveys that parsed info along with vendor attestation payload all signed by a Linux key. I.e. still allow for a unified output format + signed vendor blob and provide a path to keep all the vendor specific handling internal to the kernel.
All the specific fields of the blob have to be decoded and subjected to an acceptance policy. That policy will most always be different across different platforms and VM owners. I wrote all of github.com/google/go-sev-guest, including the verification and validation logic, and it's going to get more complicated, and the sources of the data that provide validators with notions of what values can be trusted will be varied. The formats are not standardized. The Confidential Computing Consortium should be working toward that, but it's a slow process. There's IETF RATS. There's in-toto.io attestations. There's Azure's JWT thing. There's a signed serialized protocol buffer that I've decided is what Google is going to produce while we figure out all the "right" formats to use. There will be factions and absolute gridlock for multiple years if we require solidifying an abstraction for the kernel to manage all this logic before passing a report on to user space.
Now, not only are the field contents important, the certificates of the keys that signed the report are important. Each platform has its own special x509v3 extensions and key hierarchy to express what parts of the report should be what value if signed by this key, and in TDX's case there are extra endpoints that you need to query to determine if there's an active CVE on the associated TCB version. This is how they avoid adding every cpu's key to the leaf certificate's CRL.
You really shouldn't be putting attestation validation logic in the kernel. It belongs outside of the VM entirely with the party that will only release access keys to the VM if it can prove it's running the software it claims, on the platform it claims. I think Windows puts a remote procedure call in their guest attestation driver to the Azure attestation service, and that is an anti-pattern in my mind.
ARM's Confidential Computing Realm Management Extensions (RME) seems to be going along the lines of a runtime measurement register model with their hardware enforced security. The number of registers isn't prescribed in the spec.
+Joey Gouly +linux-coco@lists.linux.dev as far as RME is concerned, do you know who would be best to weigh in on this discussion of a unified attestation model?
On Tue, 2023-06-27 at 18:36 -0700, Dionna Amalie Glaze wrote:
On Tue, Jun 27, 2023 at 5:13 PM Dan Williams dan.j.williams@intel.com wrote:
[..]
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology.
That's where I hope the line can be drawn, i.e. that all of this vendor differentiation really only matters inside the kernel in the end.
I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
This is where I need help. Can you identify where the following assertion falls over:
"The minimum viable key-server is one that can generically validate a blob with an ECDSA signature".
I.e. the fact that SEV and TDX send different length blobs is less important than validating that signature.
If it is always the case that specific fields in the blob need to be decoded then yes, that weakens the assertion. However, maybe that means that kernel code parses the blob and conveys that parsed info along with vendor attestation payload all signed by a Linux key. I.e. still allow for a unified output format + signed vendor blob and provide a path to keep all the vendor specific handling internal to the kernel.
All the specific fields of the blob have to be decoded and subjected to an acceptance policy. That policy will most always be different across different platforms and VM owners. I wrote all of github.com/google/go-sev-guest, including the verification and validation logic, and it's going to get more complicated, and the sources of the data that provide validators with notions of what values can be trusted will be varied. The formats are not standardized. The Confidential Computing Consortium should be working toward that, but it's a slow process. There's IETF RATS. There's in-toto.io attestations. There's Azure's JWT thing. There's a signed serialized protocol buffer that I've decided is what Google is going to produce while we figure out all the "right" formats to use. There will be factions and absolute gridlock for multiple years if we require solidifying an abstraction for the kernel to manage all this logic before passing a report on to user space.
Now, not only are the field contents important, the certificates of the keys that signed the report are important. Each platform has its own special x509v3 extensions and key hierarchy to express what parts of the report should be what value if signed by this key, and in TDX's case there are extra endpoints that you need to query to determine if there's an active CVE on the associated TCB version. This is how they avoid adding every cpu's key to the leaf certificate's CRL.
You really shouldn't be putting attestation validation logic in the kernel.
Agreed. The data blob for remote verification should be just some data blob to the kernel. I think the kernel shouldn't even try to understand the data blob is for which architecture. From the kernel's perspective, it should be just some data blob that the kernel gets from hardware/firmware or whatever embedded in the root-of-trust in the hardware after taking some input from usrspace for the unique identity of the blob that can be used to, e.g., mitigate replay- attack, etc.
On Wed, Jun 28, 2023 at 02:16:45AM +0000, Huang, Kai wrote:
You really shouldn't be putting attestation validation logic in the kernel.
Agreed. The data blob for remote verification should be just some data blob to the kernel. I think the kernel shouldn't even try to understand the data blob is for which architecture. From the kernel's perspective, it should be just some data blob that the kernel gets from hardware/firmware or whatever embedded in the root-of-trust in the hardware after taking some input from usrspace for the unique identity of the blob that can be used to, e.g., mitigate replay- attack, etc.
Great, then use the common "data blob" api that we have in the kernel for a very long time now, the "firwmare download" api, or the sysfs binary file api. Both of them just use the kernel as a pass-through and do not touch the data at all. No need for crazy custom ioctls and all that mess :)
thanks,
greg k-h
On Wed, 2023-06-28 at 08:46 +0200, gregkh@linuxfoundation.org wrote:
On Wed, Jun 28, 2023 at 02:16:45AM +0000, Huang, Kai wrote:
You really shouldn't be putting attestation validation logic in the kernel.
Agreed. The data blob for remote verification should be just some data blob to the kernel. I think the kernel shouldn't even try to understand the data blob is for which architecture. From the kernel's perspective, it should be just some data blob that the kernel gets from hardware/firmware or whatever embedded in the root-of-trust in the hardware after taking some input from usrspace for the unique identity of the blob that can be used to, e.g., mitigate replay- attack, etc.
Great, then use the common "data blob" api that we have in the kernel for a very long time now, the "firwmare download" api, or the sysfs binary file api. Both of them just use the kernel as a pass-through and do not touch the data at all. No need for crazy custom ioctls and all that mess :)
I guess I was talking about from "kernel shouldn't try to parse attestation data blob" perspective. Looking at AMD's attestation flow (I have no deep understanding of AMD's attestation flow), the assumption of "one remote verifiable data blob" isn't even true -- AMD can return "attestation report" (remote verifiable) and the "certificate" to verify it separately:
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/snp-attestation.html
On the other hand, AFAICT Intel SGX-based attestation doesn't have a mechanism "for the kernel" to return certificate(s), but choose to embed the certificate(s) to the Quote itself. I believe we can add such mechanism (e.g., another TDVMCALL) for the kernel to get certificate(s) separately, but AFAICT it doesn't exist yet.
Btw, getting "remote verifiable blob" is only one step of the attestation flow. For instance, before the blob can be generated, there must be a step to establish the attestation key between the machine and the attestation service. And the flow to do this could be very different between vendors too.
That being said, while I believe all those differences can be unified in some way, I think the question is whether it is worth to put such effort to try to unify attestation flow for all vendors. As Erdem Aktas mentioned earlier, "the number of CPU vendors for confidential computing seems manageable".
On Wed, Jun 28, 2023 at 08:56:30AM +0000, Huang, Kai wrote:
On Wed, 2023-06-28 at 08:46 +0200, gregkh@linuxfoundation.org wrote:
On Wed, Jun 28, 2023 at 02:16:45AM +0000, Huang, Kai wrote:
You really shouldn't be putting attestation validation logic in the kernel.
Agreed. The data blob for remote verification should be just some data blob to the kernel. I think the kernel shouldn't even try to understand the data blob is for which architecture. From the kernel's perspective, it should be just some data blob that the kernel gets from hardware/firmware or whatever embedded in the root-of-trust in the hardware after taking some input from usrspace for the unique identity of the blob that can be used to, e.g., mitigate replay- attack, etc.
Great, then use the common "data blob" api that we have in the kernel for a very long time now, the "firwmare download" api, or the sysfs binary file api. Both of them just use the kernel as a pass-through and do not touch the data at all. No need for crazy custom ioctls and all that mess :)
I guess I was talking about from "kernel shouldn't try to parse attestation data blob" perspective. Looking at AMD's attestation flow (I have no deep understanding of AMD's attestation flow), the assumption of "one remote verifiable data blob" isn't even true -- AMD can return "attestation report" (remote verifiable) and the "certificate" to verify it separately:
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/snp-attestation.html
On the other hand, AFAICT Intel SGX-based attestation doesn't have a mechanism "for the kernel" to return certificate(s), but choose to embed the certificate(s) to the Quote itself. I believe we can add such mechanism (e.g., another TDVMCALL) for the kernel to get certificate(s) separately, but AFAICT it doesn't exist yet.
Btw, getting "remote verifiable blob" is only one step of the attestation flow. For instance, before the blob can be generated, there must be a step to establish the attestation key between the machine and the attestation service. And the flow to do this could be very different between vendors too.
That being said, while I believe all those differences can be unified in some way, I think the question is whether it is worth to put such effort to try to unify attestation flow for all vendors. As Erdem Aktas mentioned earlier, "the number of CPU vendors for confidential computing seems manageable".
So you think that there should be a custom user/kernel api for every single different CPU vendor? That's not how kernel development works, sorry. Let's try to unify them to make both the kernel, and userspace, sane.
And Dan is right, if this is handling keys, then the key subsystem needs to be used here instead of custom ioctls.
thanks,
greg k-h
On Wed, 2023-06-28 at 11:02 +0200, gregkh@linuxfoundation.org wrote:
On Wed, Jun 28, 2023 at 08:56:30AM +0000, Huang, Kai wrote:
On Wed, 2023-06-28 at 08:46 +0200, gregkh@linuxfoundation.org wrote:
On Wed, Jun 28, 2023 at 02:16:45AM +0000, Huang, Kai wrote:
You really shouldn't be putting attestation validation logic in the kernel.
Agreed. The data blob for remote verification should be just some data blob to the kernel. I think the kernel shouldn't even try to understand the data blob is for which architecture. From the kernel's perspective, it should be just some data blob that the kernel gets from hardware/firmware or whatever embedded in the root-of-trust in the hardware after taking some input from usrspace for the unique identity of the blob that can be used to, e.g., mitigate replay- attack, etc.
Great, then use the common "data blob" api that we have in the kernel for a very long time now, the "firwmare download" api, or the sysfs binary file api. Both of them just use the kernel as a pass-through and do not touch the data at all. No need for crazy custom ioctls and all that mess :)
I guess I was talking about from "kernel shouldn't try to parse attestation data blob" perspective. Looking at AMD's attestation flow (I have no deep understanding of AMD's attestation flow), the assumption of "one remote verifiable data blob" isn't even true -- AMD can return "attestation report" (remote verifiable) and the "certificate" to verify it separately:
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/snp-attestation.html
On the other hand, AFAICT Intel SGX-based attestation doesn't have a mechanism "for the kernel" to return certificate(s), but choose to embed the certificate(s) to the Quote itself. I believe we can add such mechanism (e.g., another TDVMCALL) for the kernel to get certificate(s) separately, but AFAICT it doesn't exist yet.
Btw, getting "remote verifiable blob" is only one step of the attestation flow. For instance, before the blob can be generated, there must be a step to establish the attestation key between the machine and the attestation service. And the flow to do this could be very different between vendors too.
That being said, while I believe all those differences can be unified in some way, I think the question is whether it is worth to put such effort to try to unify attestation flow for all vendors. As Erdem Aktas mentioned earlier, "the number of CPU vendors for confidential computing seems manageable".
So you think that there should be a custom user/kernel api for every single different CPU vendor? That's not how kernel development works, sorry. Let's try to unify them to make both the kernel, and userspace, sane.
And Dan is right, if this is handling keys, then the key subsystem needs to be used here instead of custom ioctls.
Sure. I have no objection to this.
Dionna Amalie Glaze wrote:
On Tue, Jun 27, 2023 at 5:13 PM Dan Williams dan.j.williams@intel.com wrote:
[..]
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology.
That's where I hope the line can be drawn, i.e. that all of this vendor differentiation really only matters inside the kernel in the end.
I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
This is where I need help. Can you identify where the following assertion falls over:
"The minimum viable key-server is one that can generically validate a blob with an ECDSA signature".
I.e. the fact that SEV and TDX send different length blobs is less important than validating that signature.
If it is always the case that specific fields in the blob need to be decoded then yes, that weakens the assertion. However, maybe that means that kernel code parses the blob and conveys that parsed info along with vendor attestation payload all signed by a Linux key. I.e. still allow for a unified output format + signed vendor blob and provide a path to keep all the vendor specific handling internal to the kernel.
First, thank you for engaging, it speeds up the iteration. This confirmed my worry that the secondary goal of this proposal, a common verification implementation, is indeed unachievable in the near term. A few clarifying questions below, but I will let this go.
The primary goal, achievable on a short runway, is more for kernel developers. It is to have a common infrastructure for marshaling vendor payloads, provide a mechanism to facilitate kernel initiated requests to a key-server, and to deploy a common frontend for concepts like runtime measurement (likely as another backend to what Keys already understands for various TPM PCR implementations).
All the specific fields of the blob have to be decoded and subjected to an acceptance policy. That policy will most always be different across different platforms and VM owners. I wrote all of github.com/google/go-sev-guest, including the verification and validation logic, and it's going to get more complicated, and the sources of the data that provide validators with notions of what values can be trusted will be varied.
Can you provide an example? I ask only to include it in the kernel commit log for a crisp explanation why this proposed Keys format will continue to convey a raw vendor blob with no kernel abstraction as part of its payload for the foreseeable future.
The formats are not standardized. The Confidential Computing Consortium should be working toward that, but it's a slow process. There's IETF RATS. There's in-toto.io attestations. There's Azure's JWT thing. There's a signed serialized protocol buffer that I've decided is what Google is going to produce while we figure out all the "right" formats to use. There will be factions and absolute gridlock for multiple years if we require solidifying an abstraction for the kernel to manage all this logic before passing a report on to user space.
Understood. When that standardization process completes my expectation is that it slots into the common conveyance method and no need to go rewrite code that already knows how to interface with Keys to get attestation evidence.
Now, not only are the field contents important, the certificates of the keys that signed the report are important. Each platform has its own special x509v3 extensions and key hierarchy to express what parts of the report should be what value if signed by this key, and in TDX's case there are extra endpoints that you need to query to determine if there's an active CVE on the associated TCB version. This is how they avoid adding every cpu's key to the leaf certificate's CRL.
You really shouldn't be putting attestation validation logic in the kernel.
It was less putting validation logic in the kernel, and more hoping for a way to abstract some common parsing in advance of a true standard attestation format, but point taken.
It belongs outside of the VM entirely with the party that will only release access keys to the VM if it can prove it's running the software it claims, on the platform it claims. I think Windows puts a remote procedure call in their guest attestation driver to the Azure attestation service, and that is an anti-pattern in my mind.
I can not speak to the Windows implementation, but the Linux Keys subsystem is there to handle Key construction that may be requested by userspace or the kernel and may be serviced by built-in keys, device/platform instantiated keys, or keys retrieved via an upcall to userspace.
The observation is that existing calls to request_key() in the kernel likely have reason to be serviced by a confidential computing key server somewhere in the chain. So, might as well enlighten the Keys subsystem to retrieve this information and skip round trips to userspace run vendor specific ioctls. Make the kernel as self sufficient as possible, and make SEV, TDX, etc. developers talk more to each other about their needs.
First, thank you for engaging, it speeds up the iteration. This confirmed my worry that the secondary goal of this proposal, a common verification implementation, is indeed unachievable in the near term. A few clarifying questions below, but I will let this go.
The primary goal, achievable on a short runway, is more for kernel developers. It is to have a common infrastructure for marshaling vendor payloads, provide a mechanism to facilitate kernel initiated requests to a key-server, and to deploy a common frontend for concepts like runtime measurement (likely as another backend to what Keys already understands for various TPM PCR implementations).
That sounds good, though the devil is in the details. The TPM situation will be exacerbated by a lower root of trust. The TPM itself doesn't have a specification for its own attested firmware.
All the specific fields of the blob have to be decoded and subjected to an acceptance policy. That policy will most always be different across different platforms and VM owners. I wrote all of github.com/google/go-sev-guest, including the verification and validation logic, and it's going to get more complicated, and the sources of the data that provide validators with notions of what values can be trusted will be varied.
Can you provide an example? I ask only to include it in the kernel commit log for a crisp explanation why this proposed Keys format will continue to convey a raw vendor blob with no kernel abstraction as part of its payload for the foreseeable future.
An example is that while there is a common notion that each report will have some attestation key whose certificate needs to be verified, there is additional collateral that must be downloaded to
* verify a TDX key certificate against updates to known weaknesses of the key's details * verify the measurement in the report against a vendor's signed golden measurement * [usually offline and signed by the analyzing principal that the analysis was done] fully verify the measurement given a build provenance document like SLSA. The complexity of this analysis could even engage in static analysis of every commit since a certain date, or from a developer of low repute... whatever the verifier wants to do.
These are all in the realm of interpreting the blob for acceptance, so it's best to keep uninterpreted.
The formats are not standardized. The Confidential Computing Consortium should be working toward that, but it's a slow process. There's IETF RATS. There's in-toto.io attestations. There's Azure's JWT thing. There's a signed serialized protocol buffer that I've decided is what Google is going to produce while we figure out all the "right" formats to use. There will be factions and absolute gridlock for multiple years if we require solidifying an abstraction for the kernel to manage all this logic before passing a report on to user space.
Understood. When that standardization process completes my expectation is that it slots into the common conveyance method and no need to go rewrite code that already knows how to interface with Keys to get attestation evidence.
I can get on board with that. I don't think there will be much cause for more than a handful of attestation requests with different report data, so it shouldn't overwhelm the key subsystem.
You really shouldn't be putting attestation validation logic in the kernel.
It was less putting validation logic in the kernel, and more hoping for a way to abstract some common parsing in advance of a true standard attestation format, but point taken.
I think we'll have hardware-provided blobs and host-provided cached collateral. The caching could be achieved with a hosted proxy server, but given SEV-SNP already has GET_EXT_GUEST_REQUEST to simplify delivery, I think it's fair to offer other technologies the chance at supporting a similar simple solution.
Everything else will have to come from the network or the workload itself.
It belongs outside of the VM entirely with the party that will only release access keys to the VM if it can prove it's running the software it claims, on the platform it claims. I think Windows puts a remote procedure call in their guest attestation driver to the Azure attestation service, and that is an anti-pattern in my mind.
I can not speak to the Windows implementation, but the Linux Keys subsystem is there to handle Key construction that may be requested by userspace or the kernel and may be serviced by built-in keys, device/platform instantiated keys, or keys retrieved via an upcall to userspace.
The observation is that existing calls to request_key() in the kernel likely have reason to be serviced by a confidential computing key server somewhere in the chain. So, might as well enlighten the Keys subsystem to retrieve this information and skip round trips to userspace run vendor specific ioctls. Make the kernel as self sufficient as possible, and make SEV, TDX, etc. developers talk more to each other about their needs.
That sounds reasonable. I think one wrinkle in the current design is that SGX and SEV-SNP provide derived keys as a thing separate from attestation but still based on firmware measurement, and TDX doesn't yet. It may in the future come with a TDX module update that gets derived keys through an SGX enclave–who knows. The MSG_KEY_REQ guest request for a SEV-SNP derived key has some bits and bobs to select different VM material to mix into the key derivation, so that would need to be in the API as well. It makes request_key a little weird to use for both. I don't even think there's a sufficient abstraction for the guest-attest device to provide, since there isn't a common REPORT_DATA + attestation level pair of inputs that drive it. If we're fine with a technology-tagged uninterpreted input blob for key derivation, and the device returns an error if the real hardware doesn't match the technology tag, then that could be an okay enough interface.
I could be convinced to leave MSG_KEY_REQ out of Linux entirely, but only for selfish reasons. The alternative is to set up a sealing key escrow service that releases sealing keys when a VM's attestation matches a pre-registered policy, which is extremely heavy-handed when you can enforce workload identity at VM launch time and have a safe derived key with this technology. I think there's a Decentriq blog post about setting that whole supply chain up ("swiss cheese to cheddar"), so they'd likely have some words about that.
On Tue, Jun 27, 2023 at 06:36:07PM -0700, Dionna Amalie Glaze wrote:
On Tue, Jun 27, 2023 at 5:13 PM Dan Williams dan.j.williams@intel.com wrote:
[..]
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology.
That's where I hope the line can be drawn, i.e. that all of this vendor differentiation really only matters inside the kernel in the end.
I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
This is where I need help. Can you identify where the following assertion falls over:
"The minimum viable key-server is one that can generically validate a blob with an ECDSA signature".
I.e. the fact that SEV and TDX send different length blobs is less important than validating that signature.
If it is always the case that specific fields in the blob need to be decoded then yes, that weakens the assertion. However, maybe that means that kernel code parses the blob and conveys that parsed info along with vendor attestation payload all signed by a Linux key. I.e. still allow for a unified output format + signed vendor blob and provide a path to keep all the vendor specific handling internal to the kernel.
All the specific fields of the blob have to be decoded and subjected to an acceptance policy. That policy will most always be different across different platforms and VM owners. I wrote all of github.com/google/go-sev-guest, including the verification and validation logic, and it's going to get more complicated, and the sources of the data that provide validators with notions of what values can be trusted will be varied. The formats are not standardized. The Confidential Computing Consortium should be working toward that, but it's a slow process. There's IETF RATS. There's in-toto.io attestations. There's Azure's JWT thing. There's a signed serialized protocol buffer that I've decided is what Google is going to produce while we figure out all the "right" formats to use. There will be factions and absolute gridlock for multiple years if we require solidifying an abstraction for the kernel to manage all this logic before passing a report on to user space.
I agree with most of the above, but all that nightmate^Wcomplexity is handled on the remote attestation side. If I understand the current discussion, it's about how to abstract a guest attestation evidence generation request in a vendor agnostic way. And I think what's proposed here is simply to pass a binary payload (The evidence request from the guest userspace) to the kernel key subsystem, hook it into vendor specific handler and get userspace an attestation evidence (a platform key signed blob) back to the guest app. The guest app can then give that to an attestation service, and that's when all the above described complexity takes place. Am I missing something?
Now, not only are the field contents important, the certificates of the keys that signed the report are important. Each platform has its own special x509v3 extensions and key hierarchy to express what parts of the report should be what value if signed by this key, and in TDX's case there are extra endpoints that you need to query to determine if there's an active CVE on the associated TCB version. This is how they avoid adding every cpu's key to the leaf certificate's CRL.
You really shouldn't be putting attestation validation logic in the kernel.
AFAIU, that's not part of the proposal/PoC/mockup. It's all about funneling an attestation evidence request down to the TSM/PSP/firmware for it to generate an actually verifiable attestation evidence.
Cheers, Samuel.
On Tue, Jun 27, 2023 at 05:11:07PM -0700, Dan Williams wrote:
Dionna Amalie Glaze wrote:
On Sun, Jun 25, 2023 at 8:06 PM Sathyanarayanan Kuppuswamy sathyanarayanan.kuppuswamy@linux.intel.com wrote:
[..]
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The VMPL-based separation that will house the supervisor module known as SVSM can have protocols that implement a TPM command interface, or an RTMR-extension interface, and will also need to have an SVSM-specific protocol attestation report format to keep the secure chain of custody apparent. We'd have different formats and protocols in the kernel, at least, to speak to each technology.
That's where I hope the line can be drawn, i.e. that all of this vendor differentiation really only matters inside the kernel in the end.
Looking at your keys subsystem based PoC (thanks for putting it together), I understand that the intention is to pass an attestation evidence request as a payload to the kernel, in a abstract way. i.e. the void *data in: static int guest_attest_request_key(struct key *authkey, void *data)
And then passiing that down to vendor specific handlers (tdx_request_attest in your PoC) for it to behave as a key server for that attestation evidence request. The vendor magic of transforming an attestation request into an actual attestation evidence (typically signed with platform derived keys) is stuffed into that handler. The format, content and protection of both the attestation evidence request and the evidence itself is left for the guest kernel handler (e.g. tdx_request_attest) to handle.
Is that a fair description of your proposal?
If it is, then it makes sense to me, and could serve as a generic abstraction for confidential computing guest attestation evidence requests. I think it could support the TDX, SEV and also the RISC-V (aka CoVE) guest attestation request evidence flow.
I'm not sure it's worth the trouble of papering over all the... 3-4 technologies with similar but still weirdly different formats and ways of doing things with an abstracted attestation ABI, especially since the output all has to be interpreted in an architecture-specific way anyway.
This is where I need help. Can you identify where the following assertion falls over:
"The minimum viable key-server is one that can generically validate a blob with an ECDSA signature".
I.e. the fact that SEV and TDX send different length blobs is less important than validating that signature.
I'm not sure which signature we're talking about here. The final attestation evidence (The blob the guest workload will send to a remote attestation service) is signed with a platform derived key, typically rooted into a manufacturer's CA. It is then up to the *remote* attestation service to authenticate and validate the evidence signature. Then it can go through the actual attestation verification flow (comparison against reference values, policy checks, etc). The latter should be none of the guest kernel's business, which is what your proposal seems to be heading to.
If it is always the case that specific fields in the blob need to be decoded then yes, that weakens the assertion.
Your vendor specific key handler may have to decode the passed void pointer into a vendor specific structure before sending it down to the TSM/ASP/etc, and that's fine imho.
Cheers, Samuel.
Sathyanarayanan Kuppuswamy wrote:
Hi Dan,
On 6/23/23 3:27 PM, Dan Williams wrote:
Dan Williams wrote:
[ add David, Brijesh, and Atish]
Kuppuswamy Sathyanarayanan wrote:
In TDX guest, the second stage of the attestation process is Quote generation. This process is required to convert the locally generated TDREPORT into a remotely verifiable Quote. It involves sending the TDREPORT data to a Quoting Enclave (QE) which will verify the integrity of the TDREPORT and sign it with an attestation key.
Intel's TDX attestation driver exposes TDX_CMD_GET_QUOTE IOCTL to allow the user agent to get the TD Quote.
Add a kernel selftest module to verify the Quote generation feature.
TD Quote generation involves following steps:
- Get the TDREPORT data using TDX_CMD_GET_REPORT IOCTL.
- Embed the TDREPORT data in quote buffer and request for quote generation via TDX_CMD_GET_QUOTE IOCTL request.
- Upon completion of the GetQuote request, check for non zero value in the status field of Quote header to make sure the generated quote is valid.
What this cover letter does not say is that this is adding another instance of the similar pattern as SNP_GET_REPORT.
Linux is best served when multiple vendors trying to do similar operations are brought together behind a common ABI. We see this in the history of wrangling SCSI vendors behind common interfaces. Now multiple confidential computing vendors trying to develop similar flows with differentiated formats where that differentiation need not leak over the ABI boundary.
[..]
Below is a rough mock up of this approach to demonstrate the direction. Again, the goal is to define an ABI that can support any vendor's arch-specific attestation method and key provisioning flows without leaking vendor-specific details, or confidential material over the user/kernel ABI.
Thanks for working on this mock code and helping out. It gives me the general idea about your proposal.
The observation is that there are a sufficient number of attestation flows available to review where Linux can define a superset ABI to contain them all. The other observation is that the implementations have features that may cross-polinate over time. For example the SEV privelege level consideration ("vmpl"), and the TDX RTMR (think TPM PCRs) mechanisms address generic Confidential Computing use cases.
I agree with your point about VMPL and RTMR feature cases. This observation is valid for AMD SEV and TDX attestation flows. But I am not sure whether it will hold true for other vendor implementations. Our sample set is not good enough to make this conclusion. The reason for my concern is, if you check the ABI interface used in the S390 arch attestation driver (drivers/s390/char/uvdevice.c), you would notice that there is a significant difference between the ABI used in that driver and SEV/TDX drivers. The S390 driver attestation request appears to accept two data blobs as input, as well as a variety of vendor-specific header configurations.
I would need more time to investigate. It's also the case that if both major x86 vendors plus ARM and/or RISC-V can all get behind the same frontend then that is already success in my mind.
Maybe the s390 attestation model is a special case, but, I think we consider this issue. Since we don't have a common spec, there is chance that any superset ABI we define now may not meet future vendor requirements. One way to handle it to leave enough space in the generic ABI to handle future vendor requirements.
Perhaps, but the goal here is to clearly indicate "this is how Linux conveys confidential computing attestation concepts". If there is future vendor innovation in this space it needs to consider how it meets the established needs of Linux, not the other way round.
I think it would be better if other vendors (like ARM or RISC) can comment and confirm whether this proposal meets their demands.
The more participation the better. Open source definitely involves a component speaking up when the definition of things are still malleable, or catching issues before they go upstream.
Vendor specific ioctls for all of this feels like surrender when Linux already has the keys subsystem which has plenty of degrees of freedom for tracking blobs with signatures and using those blobs to instantiate other blobs. It already serves as the ABI wrapping various TPM implementations and marshaling keys for storage encryption and other use cases that intersect Confidential Computing.
The benefit of deprecating vendor-specific abstraction layers in userspace is secondary. The primary benefit is collaboration. It enables kernel developers from various architectures to collaborate on common infrastructure. If, referring back to my previous example, SEV adopts an RTMR-like mechanism and TDX adopts a vmpl-like mechanism it would be unfortunate if those efforts were siloed, duplicated, and needlessly differentiated to userspace. So while there are arguably a manageable number of basic arch attestation methods the planned expansion of those to build incremental functionality is where I believe we, as a community, will be glad that we invested in a "Linux format" for all of this.
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel. Now, this is only the
IIUC, the idea here is to cache the quote data and return it to the user whenever possible, right? If yes, I think such optimization may not be very useful for our case. AFAIK, the quote data will change whenever there is a change in the guest measurement data. Since the validity of the generated quote will not be long, and the frequency of quote generation requests is expected to be less, we may not get much benefit from caching the quote data. I think we can keep this logic simple by directly retrieving the quote data from the quoting enclave whenever there is a request from the user.
The Keys subsystem already supports the concept of keys that expire immediately, so no need for special consideration here that I can see.
top-half of what is needed. The missing bottom half takes that material and uses it to instantiate derived key material like the storage decryption key internal to the kernel. See "The Process" in Documentation/security/keys/request-key.rst for how the Keys subsystem handles the "keys for keys" use case.
This is only useful for key-server use case, right? Attestation can also be used for use cases like pattern matching or uploading some secure data, etc. Since key-server is not the only use case, does it make sense to suppport this derived key feature?
The Keys subsystem is just a way for both userspace and kernel space to request the instantiation of blobs that mediate access to another resource be it another key or something else. So key-server is only one example client.
The other reason for defining a common frontend is so the kernel can understand and mediate resource access as a kernel is wont to do. The ioctl() approach blinds the kernel and requires userspace to repeatedly solve problems in vendor specific ways. They Keys proposal also has the property of not requiring userspace round trips for things like unlocking storage. The driver can just do request_keys() and kick off the process on its own.
diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig index f79ab13a5c28..0f775847028e 100644 --- a/drivers/virt/Kconfig +++ b/drivers/virt/Kconfig @@ -54,4 +54,8 @@ source "drivers/virt/coco/sev-guest/Kconfig" source "drivers/virt/coco/tdx-guest/Kconfig" +config GUEST_ATTEST
- tristate
- select KEYS
endif diff --git a/drivers/virt/Makefile b/drivers/virt/Makefile index e9aa6fc96fab..66f6b838f8f4 100644 --- a/drivers/virt/Makefile +++ b/drivers/virt/Makefile @@ -12,3 +12,4 @@ obj-$(CONFIG_ACRN_HSM) += acrn/ obj-$(CONFIG_EFI_SECRET) += coco/efi_secret/ obj-$(CONFIG_SEV_GUEST) += coco/sev-guest/ obj-$(CONFIG_INTEL_TDX_GUEST) += coco/tdx-guest/ +obj-$(CONFIG_GUEST_ATTEST) += coco/guest-attest/ diff --git a/drivers/virt/coco/guest-attest/Makefile b/drivers/virt/coco/guest-attest/Makefile new file mode 100644 index 000000000000..5581c5a27588 --- /dev/null +++ b/drivers/virt/coco/guest-attest/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_GUEST_ATTEST) += guest_attest.o +guest_attest-y := key.o diff --git a/drivers/virt/coco/guest-attest/key.c b/drivers/virt/coco/guest-attest/key.c new file mode 100644 index 000000000000..2a494b6dd7a7 --- /dev/null +++ b/drivers/virt/coco/guest-attest/key.c @@ -0,0 +1,159 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright(c) 2023 Intel Corporation. All rights reserved. */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/seq_file.h> +#include <linux/key-type.h> +#include <linux/module.h> +#include <linux/base64.h>
+#include <keys/request_key_auth-type.h> +#include <keys/user-type.h>
+#include "guest-attest.h"
Can you share you guest-attest.h?
Apologies, missed a 'git add':
/* SPDX-License-Identifier: GPL-2.0-only */ /* Copyright(c) 2023 Intel Corporation. */
#ifndef __GUEST_ATTEST_H__ #define __GUEST_ATTEST_H__ #include <linux/list.h>
/* * arch specific ops, only one is expected to be registered at a time * i.e. either SEV or TDX, but not both */ struct guest_attest_ops { const char *name; struct module *module; struct list_head list; int (*request_attest)(struct key *key, int level, struct key *dest_keyring, void *payload, int payload_len, struct key *authkey); };
#define GUEST_ATTEST_DATALEN 64
int register_guest_attest_ops(struct guest_attest_ops *ops); void unregister_guest_attest_ops(struct guest_attest_ops *ops);
#endif /*__GUEST_ATTEST_H__ */
+static LIST_HEAD(guest_attest_list); +static DECLARE_RWSEM(guest_attest_rwsem);
+static struct guest_attest_ops *fetch_ops(void) +{
- return list_first_entry_or_null(&guest_attest_list,
struct guest_attest_ops, list);+}
+static struct guest_attest_ops *get_ops(void) +{
- down_read(&guest_attest_rwsem);
- return fetch_ops();
+}
+static void put_ops(void) +{
- up_read(&guest_attest_rwsem);
+}
+int register_guest_attest_ops(struct guest_attest_ops *ops) +{
- struct guest_attest_ops *conflict;
- int rc;
- down_write(&guest_attest_rwsem);
- conflict = fetch_ops();
- if (conflict) {
pr_err("\"%s\" ops already registered\n", conflict->name);rc = -EEXIST;goto out;- }
- list_add(&ops->list, &guest_attest_list);
- try_module_get(ops->module);
- rc = 0;
+out:
- up_write(&guest_attest_rwsem);
- return rc;
+} +EXPORT_SYMBOL_GPL(register_guest_attest_ops);
+void unregister_guest_attest_ops(struct guest_attest_ops *ops) +{
- down_write(&guest_attest_rwsem);
- list_del(&ops->list);
- up_write(&guest_attest_rwsem);
- module_put(ops->module);
+} +EXPORT_SYMBOL_GPL(unregister_guest_attest_ops);
+static int __guest_attest_request_key(struct key *key, int level,
struct key *dest_keyring,const char *callout_info, int callout_len,struct key *authkey)+{
- struct guest_attest_ops *ops;
- void *payload = NULL;
- int rc, payload_len;
- ops = get_ops();
- if (!ops)
return -ENOKEY;- payload = kzalloc(max(GUEST_ATTEST_DATALEN, callout_len), GFP_KERNEL);
- if (!payload) {
rc = -ENOMEM;goto out;- }
Is the idea to get the values like vmpl part of the payload?
No, to me vmpl likely needs to be conveyed in the key-description. Payload is simply the 64-bytes that both SEV and TDX take as input for the attestation request. The AMD specification seems to imply that payload is itself a public key? If that is the expectation then it may be more appropriate to make that a separate retrieval vs something passed in directly from userspace.
- payload_len = base64_decode(callout_info, callout_len, payload);
- if (payload_len < 0 || payload_len > GUEST_ATTEST_DATALEN) {
rc = -EINVAL;goto out;- }
- rc = ops->request_attest(key, level, dest_keyring, payload, payload_len,
authkey);+out:
- kfree(payload);
- put_ops();
- return rc;
+}
+static int guest_attest_request_key(struct key *authkey, void *data) +{
- struct request_key_auth *rka = get_request_key_auth(authkey);
- struct key *key = rka->target_key;
- unsigned long long id;
- int rc, level;
- pr_debug("desc: %s op: %s callout: %s\n", key->description, rka->op,
rka->callout_info ? (char *)rka->callout_info : "\"none\"");- if (sscanf(key->description, "guest_attest:%d:%llu", &level, &id) != 2)
return -EINVAL;Can you explain some details about the id and level? It is not very clear why we need it.
@level is my mockup of the vmpl concept, and @id is something free-form for now that the requester of the attestation knows what it means. This id-scheme would follow a Linux-defined format. More discussion is needed here about what attestation data is used for and opportunities to define a common naming scheme. For example, what is the common name of the guest_attest-key that supports retrieving the storage decryption key from the key-server.
On Fri, 2023-06-23 at 15:27 -0700, Dan Williams wrote:
An example, to show what the strawman patch below enables: (req_key is the sample program from "man 2 request_key")
# ./req_key guest_attest guest_attest:0:0-$desc $(cat user_data | base64) Key ID is 10e2f3a7 # keyctl pipe 0x10e2f3a7 | hexdump -C 00000000 54 44 58 20 47 65 6e 65 72 61 74 65 64 20 51 75 |TDX Generated Qu| 00000010 6f 74 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |ote.............| 00000020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00004000
This is the kernel instantiating a TDX Quote without the TDREPORT implementation detail ever leaving the kernel.
There might be one small issue here. The generated Quote has a userspace provided 'u64 REPORTDATA' (which originally comes from userspace when generating the TDREPORT) which is supposed to be used by the attestation service to uniquely identify this Quote to mitigate some sort of reply-attack. For instance, the REPORTDATA could be a per-TLS-session data provided by the attestation service.
I don't know whether other archs have similar thing in their Quote-like blob, but I believe this in general is a reasonable thing.
IIUC, one problem of using above request_key() to generate the Quote is potentially other userspace processes are able to see this, while I believe this REPORTDATA is only supposed to be visible by the application which is responsible for talking to the attestation service.
I am not sure whether this is a risk, but using IOCTL() should be able to avoid this risk.
Tested-by: Qinkun Bao qinkun@google.com
Thanks Sathyanarayanan for the new patch! This patch is critical for our use case. We built a guest image with the patch, and verified it works for us, when using a host kernel built with https://github.com/intel/tdx repo.
On Sun, May 14, 2023 at 12:24 AM Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
Following are the details of the patch set:
Patch 1/3 -> Adds event notification IRQ support. Patch 2/3 -> Adds Quote generation support. Patch 3/3 -> Adds selftest support for Quote generation feature.
[1] https://cdrdv2.intel.com/v1/dl/getContent/726790, section titled "TDG.VP.VMCALL<GetQuote>". [2] https://lore.kernel.org/lkml/CAAYXXYxxs2zy_978GJDwKfX5Hud503gPc8=1kQ-+JwG_kA... [3] https://lore.kernel.org/lkml/a69faebb-11e8-b386-d591-dbd08330b008@linux.alib...
Kuppuswamy Sathyanarayanan (3): x86/tdx: Add TDX Guest event notify interrupt support virt: tdx-guest: Add Quote generation support selftests/tdx: Test GetQuote TDX attestation feature
Documentation/virt/coco/tdx-guest.rst | 11 ++ arch/x86/coco/tdx/tdx.c | 194 +++++++++++++++++++ arch/x86/include/asm/tdx.h | 8 + drivers/virt/coco/tdx-guest/tdx-guest.c | 175 ++++++++++++++++- include/uapi/linux/tdx-guest.h | 44 +++++ tools/testing/selftests/tdx/tdx_guest_test.c | 65 ++++++- 6 files changed, 490 insertions(+), 7 deletions(-)
-- 2.34.1
Hi,
On 5/24/23 2:33 PM, Chong Cai wrote:
Tested-by: Qinkun Bao qinkun@google.com
Thanks Sathyanarayanan for the new patch! This patch is critical for our use case. We built a guest image with the patch, and verified it works for us, when using a host kernel built with https://github.com/intel/tdx repo.
Qinkun Bao/Chong Cai, thanks for testing it. I really appreciate the help.
Dave/Boris, could you please take a look at this patch set?
On Sun, May 14, 2023 at 12:24 AM Kuppuswamy Sathyanarayanan sathyanarayanan.kuppuswamy@linux.intel.com wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
Following are the details of the patch set:
Patch 1/3 -> Adds event notification IRQ support. Patch 2/3 -> Adds Quote generation support. Patch 3/3 -> Adds selftest support for Quote generation feature.
[1] https://cdrdv2.intel.com/v1/dl/getContent/726790, section titled "TDG.VP.VMCALL<GetQuote>". [2] https://lore.kernel.org/lkml/CAAYXXYxxs2zy_978GJDwKfX5Hud503gPc8=1kQ-+JwG_kA... [3] https://lore.kernel.org/lkml/a69faebb-11e8-b386-d591-dbd08330b008@linux.alib...
Kuppuswamy Sathyanarayanan (3): x86/tdx: Add TDX Guest event notify interrupt support virt: tdx-guest: Add Quote generation support selftests/tdx: Test GetQuote TDX attestation feature
Documentation/virt/coco/tdx-guest.rst | 11 ++ arch/x86/coco/tdx/tdx.c | 194 +++++++++++++++++++ arch/x86/include/asm/tdx.h | 8 + drivers/virt/coco/tdx-guest/tdx-guest.c | 175 ++++++++++++++++- include/uapi/linux/tdx-guest.h | 44 +++++ tools/testing/selftests/tdx/tdx_guest_test.c | 65 ++++++- 6 files changed, 490 insertions(+), 7 deletions(-)
-- 2.34.1
Kuppuswamy Sathyanarayanan wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
I missed that the TDREPORT ioctl() and this character device are already upstream. The TDREPORT ioctl() if it is only needed for quote generation seems a waste because it just retrieves a blob that needs to be turned around and injected back into the kernel to generate a quote.
An ABI wants to care about the abstractions around what the hardware mechanism enables. The TD quote is not even at the end of that chain of what the ABI needs to offer. The guest wants to use the TD quote to access / unlock other resources, just like the SEV report is used to "...provide the VM with secrets, such as a disk decryption key, or other keys required for operation".
That's where the ABI line needs to be drawn. I.e. for the guest to be able to request the distributions of keys to unlock resources by a key-type and key-descriptor. Enable userspace to interrogate an attestation object without blobs needing to traverse the kernel. If the remote service needs more than just a blob and signature to validate the state of the guest then provide faclity to interrogate that property of quote / report in a common way versus the ABI risk of conveying vendor specific binary data formats in the kernel ABI.
Dan Williams wrote:
Kuppuswamy Sathyanarayanan wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
I missed that the TDREPORT ioctl() and this character device are already upstream. The TDREPORT ioctl() if it is only needed for quote generation seems a waste because it just retrieves a blob that needs to be turned around and injected back into the kernel to generate a quote.
An ABI wants to care about the abstractions around what the hardware mechanism enables. The TD quote is not even at the end of that chain of what the ABI needs to offer. The guest wants to use the TD quote to access / unlock other resources, just like the SEV report is used to "...provide the VM with secrets, such as a disk decryption key, or other keys required for operation".
That's where the ABI line needs to be drawn. I.e. for the guest to be able to request the distributions of keys to unlock resources by a key-type and key-descriptor. Enable userspace to interrogate an attestation object without blobs needing to traverse the kernel. If the remote service needs more than just a blob and signature to validate the state of the guest then provide faclity to interrogate that property of quote / report in a common way versus the ABI risk of conveying vendor specific binary data formats in the kernel ABI.
A proposal for how this space moves forward:
1/ Stop accepting new arch specific ioctls in this space and revert the Intel TDREPORT ioctl if its only reason for existing is "quote" generation.
2/ Define a container format / envelope for platform-provided attestation evidence.
The observation here is that although it is too late to unify the evidence formats across vendors, they appear to share the common form of a blob with an ECDSA signature. That reduces the minimum viable attestation service to something that can generically verify an evidence-blob signature.
3/ Define a key-description format that considers a superset of the platform needs. For example a 'privelege-level' concept can map to 'vmpl' on AMD, but be ignored for now for Intel.
4/ For in progress enabling concepts like runtime measurement registers, look to reuse / abstract that behind the Keys subsystem existing support for managing TPM PCRs.
5/ Deprecate the multiple arch specific attestation ioctl interfaces in favor of this unified conveyance method.
On 6/23/23 9:05 PM, Dan Williams wrote:
Kuppuswamy Sathyanarayanan wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
I missed that the TDREPORT ioctl() and this character device are already upstream. The TDREPORT ioctl() if it is only needed for quote generation seems a waste because it just retrieves a blob that needs to be turned around and injected back into the kernel to generate a quote.
Although the end goal is to generate the quote, the method the user chooses to achieve it may differ for a variety of reasons. In this case, we're trying to support the use case where the user will use methods like TCP/IP or vsock to generate the Quote. They can use the GET_REPORT IOCTL to get the TDREPORT and send it to the quoting enclave via the above-mentioned methods. TDVMCALL-based quote generation is intended for users who, for a variety of security reasons, do not wish to use the methods described above.
Sathyanarayanan Kuppuswamy wrote:
On 6/23/23 9:05 PM, Dan Williams wrote:
Kuppuswamy Sathyanarayanan wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
I missed that the TDREPORT ioctl() and this character device are already upstream. The TDREPORT ioctl() if it is only needed for quote generation seems a waste because it just retrieves a blob that needs to be turned around and injected back into the kernel to generate a quote.
Although the end goal is to generate the quote, the method the user chooses to achieve it may differ for a variety of reasons. In this case, we're trying to support the use case where the user will use methods like TCP/IP or vsock to generate the Quote. They can use the GET_REPORT IOCTL to get the TDREPORT and send it to the quoting enclave via the above-mentioned methods. TDVMCALL-based quote generation is intended for users who, for a variety of security reasons, do not wish to use the methods described above.
This flexibility could be supported with keys if necessary, although I would want to hear strong reasons not a "variety of reasons" why everyone cannot use a unified approach. ABI proliferation has a maintenance cost and a collaboration cost. It is within the kernel community's right to judge the cost of ABI flexibility and opt for a constrained implementation if that cost is too high.
What I would ask of those who absolutely cannot support the TDVMCALL method is to contribute a solution that intercepts the "upcall" to the platform "guest_attest_ops" and turn it into a typical keys upcall to userspace that can use the report data with a vsock tunnel.
That way the end result is still the same, a key established with the TDX Quote evidence contained within a Linux-defined envelope.
On Sun, Jun 25, 2023 at 9:32 PM Dan Williams dan.j.williams@intel.com wrote:
Sathyanarayanan Kuppuswamy wrote:
On 6/23/23 9:05 PM, Dan Williams wrote:
Kuppuswamy Sathyanarayanan wrote:
Hi All,
In TDX guest, the attestation process is used to verify the TDX guest trustworthiness to other entities before provisioning secrets to the guest.
The TDX guest attestation process consists of two steps:
- TDREPORT generation
- Quote generation.
The First step (TDREPORT generation) involves getting the TDX guest measurement data in the format of TDREPORT which is further used to validate the authenticity of the TDX guest. The second step involves sending the TDREPORT to a Quoting Enclave (QE) server to generate a remotely verifiable Quote. TDREPORT by design can only be verified on the local platform. To support remote verification of the TDREPORT, TDX leverages Intel SGX Quoting Enclave to verify the TDREPORT locally and convert it to a remotely verifiable Quote. Although attestation software can use communication methods like TCP/IP or vsock to send the TDREPORT to QE, not all platforms support these communication models. So TDX GHCI specification [1] defines a method for Quote generation via hypercalls. Please check the discussion from Google [2] and Alibaba [3] which clarifies the need for hypercall based Quote generation support. This patch set adds this support.
Support for TDREPORT generation already exists in the TDX guest driver. This patchset extends the same driver to add the Quote generation support.
I missed that the TDREPORT ioctl() and this character device are already upstream. The TDREPORT ioctl() if it is only needed for quote generation seems a waste because it just retrieves a blob that needs to be turned around and injected back into the kernel to generate a quote.
Although the end goal is to generate the quote, the method the user chooses to achieve it may differ for a variety of reasons. In this case, we're trying to support the use case where the user will use methods like TCP/IP or vsock to generate the Quote. They can use the GET_REPORT IOCTL to get the TDREPORT and send it to the quoting enclave via the above-mentioned methods. TDVMCALL-based quote generation is intended for users who, for a variety of security reasons, do not wish to use the methods described above.
This flexibility could be supported with keys if necessary, although I would want to hear strong reasons not a "variety of reasons" why everyone cannot use a unified approach. ABI proliferation has a maintenance cost and a collaboration cost. It is within the kernel community's right to judge the cost of ABI flexibility and opt for a constrained implementation if that cost is too high.
What I would ask of those who absolutely cannot support the TDVMCALL method is to contribute a solution that intercepts the "upcall" to the platform "guest_attest_ops" and turn it into a typical keys upcall to userspace that can use the report data with a vsock tunnel.
That way the end result is still the same, a key established with the TDX Quote evidence contained within a Linux-defined envelope.
I agree a unified ABI across vendors would be ideal in the long term. However, it sounds like a non-trivial task and could take quite some time to achieve. Given there's already an AMD equivalent approach upstreamed, can we also allow this TDVMCALL patch as an intermediate step to unblock various TDX attestation user cases while targeting unified ABI? The TDVMCALL here is quite isolated and serves a very specific purpose, it should be very low risk to other kernel features and easy to be reverted in the future.
On Tue, Jun 27 2023 at 00:50, Chong Cai wrote:
On Sun, Jun 25, 2023 at 9:32 PM Dan Williams dan.j.williams@intel.com wrote:
What I would ask of those who absolutely cannot support the TDVMCALL method is to contribute a solution that intercepts the "upcall" to the platform "guest_attest_ops" and turn it into a typical keys upcall to userspace that can use the report data with a vsock tunnel.
That way the end result is still the same, a key established with the TDX Quote evidence contained within a Linux-defined envelope.
I agree a unified ABI across vendors would be ideal in the long term. However, it sounds like a non-trivial task and could take quite some time to achieve. Given there's already an AMD equivalent approach upstreamed, can we also allow this TDVMCALL patch as an intermediate step to unblock various TDX attestation user cases while targeting unified ABI? The TDVMCALL here is quite isolated and serves a very specific purpose, it should be very low risk to other kernel features and easy to be reverted in the future.
No way. This is exactly how the kernel ends up with an unmaintainable mess simply because this creates an user space ABI which is not revertable ever.
It's bad enough that nobody paid attention when the AMD muck was merged, but that does not make an argument or any form of justification to add more of this.
Dan's proposal makes a lot of sense and allows to implement this in a mostly vendor agnostic way. While the AMD interface is not going away due to that, I'm 100% confident (pun intended) that such an unified interface is going to be utilized and supported by AMD (or any other vendor) sooner than later simply because the user space people who have to implement vendor agnostic orchestration tools will go for it as it makes their life easier too.
The time wasted to argue about this TDX ioctl mess could have been spent to actually migrate TDX over to this scheme. But sure it's way simpler to flog a dead horse instead of actually sitting down and getting useful work done.
Thanks,
tglx
linux-kselftest-mirror@lists.linaro.org
-
Chong Cai -
Dan Williams -
Dionna Amalie Glaze -
Erdem Aktas -
gregkh＠linuxfoundation.org -
Huang, Kai -
Kuppuswamy Sathyanarayanan -
Samuel Ortiz -
Sathyanarayanan Kuppuswamy
-
Thomas Gleixner