From: Karol Wachowski <karol.wachowski(a)intel.com>
Trigger full device recovery when the driver fails to restore device state
via engine reset and resume operations. This is necessary because, even if
submissions from a faulty context are blocked, the NPU may still process
previously submitted faulty jobs if the engine reset fails to abort them.
Such jobs can continue to generate faults and occupy device resources.
When engine reset is ineffective, the only way to recover is to perform
a full device recovery.
Fixes: dad945c27a42 ("accel/ivpu: Add handling of VPU_JSM_STATUS_MVNCI_CONTEXT_VIOLATION_HW")
Cc: <stable(a)vger.kernel.org> # v6.15+
Signed-off-by: Karol Wachowski <karol.wachowski(a)intel.com>
Signed-off-by: Jacek Lawrynowicz <jacek.lawrynowicz(a)linux.intel.com>
---
drivers/accel/ivpu/ivpu_job.c | 6 ++++--
drivers/accel/ivpu/ivpu_jsm_msg.c | 9 +++++++--
2 files changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/accel/ivpu/ivpu_job.c b/drivers/accel/ivpu/ivpu_job.c
index 1c8e283ad9854..fae8351aa3309 100644
--- a/drivers/accel/ivpu/ivpu_job.c
+++ b/drivers/accel/ivpu/ivpu_job.c
@@ -986,7 +986,8 @@ void ivpu_context_abort_work_fn(struct work_struct *work)
return;
if (vdev->fw->sched_mode == VPU_SCHEDULING_MODE_HW)
- ivpu_jsm_reset_engine(vdev, 0);
+ if (ivpu_jsm_reset_engine(vdev, 0))
+ return;
mutex_lock(&vdev->context_list_lock);
xa_for_each(&vdev->context_xa, ctx_id, file_priv) {
@@ -1009,7 +1010,8 @@ void ivpu_context_abort_work_fn(struct work_struct *work)
if (vdev->fw->sched_mode != VPU_SCHEDULING_MODE_HW)
goto runtime_put;
- ivpu_jsm_hws_resume_engine(vdev, 0);
+ if (ivpu_jsm_hws_resume_engine(vdev, 0))
+ return;
/*
* In hardware scheduling mode NPU already has stopped processing jobs
* and won't send us any further notifications, thus we have to free job related resources
diff --git a/drivers/accel/ivpu/ivpu_jsm_msg.c b/drivers/accel/ivpu/ivpu_jsm_msg.c
index 219ab8afefabd..0256b2dfefc10 100644
--- a/drivers/accel/ivpu/ivpu_jsm_msg.c
+++ b/drivers/accel/ivpu/ivpu_jsm_msg.c
@@ -7,6 +7,7 @@
#include "ivpu_hw.h"
#include "ivpu_ipc.h"
#include "ivpu_jsm_msg.h"
+#include "ivpu_pm.h"
#include "vpu_jsm_api.h"
const char *ivpu_jsm_msg_type_to_str(enum vpu_ipc_msg_type type)
@@ -163,8 +164,10 @@ int ivpu_jsm_reset_engine(struct ivpu_device *vdev, u32 engine)
ret = ivpu_ipc_send_receive(vdev, &req, VPU_JSM_MSG_ENGINE_RESET_DONE, &resp,
VPU_IPC_CHAN_ASYNC_CMD, vdev->timeout.jsm);
- if (ret)
+ if (ret) {
ivpu_err_ratelimited(vdev, "Failed to reset engine %d: %d\n", engine, ret);
+ ivpu_pm_trigger_recovery(vdev, "Engine reset failed");
+ }
return ret;
}
@@ -354,8 +357,10 @@ int ivpu_jsm_hws_resume_engine(struct ivpu_device *vdev, u32 engine)
ret = ivpu_ipc_send_receive(vdev, &req, VPU_JSM_MSG_HWS_RESUME_ENGINE_DONE, &resp,
VPU_IPC_CHAN_ASYNC_CMD, vdev->timeout.jsm);
- if (ret)
+ if (ret) {
ivpu_err_ratelimited(vdev, "Failed to resume engine %d: %d\n", engine, ret);
+ ivpu_pm_trigger_recovery(vdev, "Engine resume failed");
+ }
return ret;
}
--
2.45.1
A potential NULL pointer dereference may occur when accessing
tmp_mqd->cp_hqd_pq_control without verifying that tmp_mqd is non-NULL.
This may happen if mqd_backup[mqd_idx] is unexpectedly NULL.
Although a NULL check for mqd_backup[mqd_idx] existed previously, it was
moved to a position after the dereference in a recent commit, which
renders it ineffective.
Add an explicit NULL check for tmp_mqd before dereferencing its members.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Cc: stable(a)vger.kernel.org # v5.13+
Fixes: a330b52a9e59 ("drm/amdgpu: Init the cp MQD if it's not be initialized before")
Signed-off-by: Alexey Nepomnyashih <sdl(a)nppct.ru>
---
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 10 ++++------
1 file changed, 4 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index d7db4cb907ae..134cab16a00d 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -3817,10 +3817,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
* check mqd->cp_hqd_pq_control since this value should not be 0
*/
tmp_mqd = (struct v9_mqd *)adev->gfx.kiq[0].mqd_backup;
- if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control){
+ if (amdgpu_in_reset(adev) && tmp_mqd && tmp_mqd->cp_hqd_pq_control) {
/* for GPU_RESET case , reset MQD to a clean status */
- if (adev->gfx.kiq[0].mqd_backup)
- memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct v9_mqd_allocation));
+ memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct v9_mqd_allocation));
/* reset ring buffer */
ring->wptr = 0;
@@ -3863,7 +3862,7 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore)
*/
tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
- if (!restore && (!tmp_mqd->cp_hqd_pq_control ||
+ if (!restore && tmp_mqd && (!tmp_mqd->cp_hqd_pq_control ||
(!amdgpu_in_reset(adev) && !adev->in_suspend))) {
memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
@@ -3874,8 +3873,7 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore)
soc15_grbm_select(adev, 0, 0, 0, 0, 0);
mutex_unlock(&adev->srbm_mutex);
- if (adev->gfx.mec.mqd_backup[mqd_idx])
- memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+ memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
} else {
/* restore MQD to a clean status */
if (adev->gfx.mec.mqd_backup[mqd_idx])
--
2.43.0
This is the second attempt at achieving the same goal. This time, the
submission avoids forking the current code base, ensuring it remains
easier to maintain over time.
The set has been tested using the SCM_RIGHTS test suite [1] using QEMU
and has been seen to successfully mitigate a UAF on on a top tier
handset.
RESULTS:
TAP version 13
1..20
# Starting 20 tests from 5 test cases.
# RUN scm_rights.dgram.self_ref ...
# OK scm_rights.dgram.self_ref
ok 1 scm_rights.dgram.self_ref
# RUN scm_rights.dgram.triangle ...
# OK scm_rights.dgram.triangle
ok 2 scm_rights.dgram.triangle
# RUN scm_rights.dgram.cross_edge ...
# OK scm_rights.dgram.cross_edge
ok 3 scm_rights.dgram.cross_edge
# RUN scm_rights.dgram.backtrack_from_scc ...
# OK scm_rights.dgram.backtrack_from_scc
ok 4 scm_rights.dgram.backtrack_from_scc
# RUN scm_rights.stream.self_ref ...
# OK scm_rights.stream.self_ref
ok 5 scm_rights.stream.self_ref
# RUN scm_rights.stream.triangle ...
# OK scm_rights.stream.triangle
ok 6 scm_rights.stream.triangle
# RUN scm_rights.stream.cross_edge ...
# OK scm_rights.stream.cross_edge
ok 7 scm_rights.stream.cross_edge
# RUN scm_rights.stream.backtrack_from_scc ...
# OK scm_rights.stream.backtrack_from_scc
ok 8 scm_rights.stream.backtrack_from_scc
# RUN scm_rights.stream_oob.self_ref ...
# OK scm_rights.stream_oob.self_ref
ok 9 scm_rights.stream_oob.self_ref
# RUN scm_rights.stream_oob.triangle ...
# OK scm_rights.stream_oob.triangle
ok 10 scm_rights.stream_oob.triangle
# RUN scm_rights.stream_oob.cross_edge ...
# OK scm_rights.stream_oob.cross_edge
ok 11 scm_rights.stream_oob.cross_edge
# RUN scm_rights.stream_oob.backtrack_from_scc ...
# OK scm_rights.stream_oob.backtrack_from_scc
ok 12 scm_rights.stream_oob.backtrack_from_scc
# RUN scm_rights.stream_listener.self_ref ...
# OK scm_rights.stream_listener.self_ref
ok 13 scm_rights.stream_listener.self_ref
# RUN scm_rights.stream_listener.triangle ...
# OK scm_rights.stream_listener.triangle
ok 14 scm_rights.stream_listener.triangle
# RUN scm_rights.stream_listener.cross_edge ...
# OK scm_rights.stream_listener.cross_edge
ok 15 scm_rights.stream_listener.cross_edge
# RUN scm_rights.stream_listener.backtrack_from_scc ...
# OK scm_rights.stream_listener.backtrack_from_scc
ok 16 scm_rights.stream_listener.backtrack_from_scc
# RUN scm_rights.stream_listener_oob.self_ref ...
# OK scm_rights.stream_listener_oob.self_ref
ok 17 scm_rights.stream_listener_oob.self_ref
# RUN scm_rights.stream_listener_oob.triangle ...
# OK scm_rights.stream_listener_oob.triangle
ok 18 scm_rights.stream_listener_oob.triangle
# RUN scm_rights.stream_listener_oob.cross_edge ...
# OK scm_rights.stream_listener_oob.cross_edge
ok 19 scm_rights.stream_listener_oob.cross_edge
# RUN scm_rights.stream_listener_oob.backtrack_from_scc ...
# OK scm_rights.stream_listener_oob.backtrack_from_scc
ok 20 scm_rights.stream_listener_oob.backtrack_from_scc
# PASSED: 20 / 20 tests passed.
# Totals: pass:20 fail:0 xfail:0 xpass:0 skip:0 error:0
[0] https://lore.kernel.org/all/20250304030149.82265-1-kuniyu@amazon.com/
[1] https://lore.kernel.org/all/20240325202425.60930-16-kuniyu@amazon.com/
Alexander Mikhalitsyn (1):
af_unix: Kconfig: make CONFIG_UNIX bool
Kuniyuki Iwashima (24):
af_unix: Return struct unix_sock from unix_get_socket().
af_unix: Run GC on only one CPU.
af_unix: Try to run GC async.
af_unix: Replace BUG_ON() with WARN_ON_ONCE().
af_unix: Remove io_uring code for GC.
af_unix: Remove CONFIG_UNIX_SCM.
af_unix: Allocate struct unix_vertex for each inflight AF_UNIX fd.
af_unix: Allocate struct unix_edge for each inflight AF_UNIX fd.
af_unix: Link struct unix_edge when queuing skb.
af_unix: Bulk update unix_tot_inflight/unix_inflight when queuing skb.
af_unix: Iterate all vertices by DFS.
af_unix: Detect Strongly Connected Components.
af_unix: Save listener for embryo socket.
af_unix: Fix up unix_edge.successor for embryo socket.
af_unix: Save O(n) setup of Tarjan's algo.
af_unix: Skip GC if no cycle exists.
af_unix: Avoid Tarjan's algorithm if unnecessary.
af_unix: Assign a unique index to SCC.
af_unix: Detect dead SCC.
af_unix: Replace garbage collection algorithm.
af_unix: Remove lock dance in unix_peek_fds().
af_unix: Try not to hold unix_gc_lock during accept().
af_unix: Don't access successor in unix_del_edges() during GC.
af_unix: Add dead flag to struct scm_fp_list.
Michal Luczaj (1):
af_unix: Fix garbage collection of embryos carrying OOB with
SCM_RIGHTS
Shigeru Yoshida (1):
af_unix: Fix uninit-value in __unix_walk_scc()
include/net/af_unix.h | 48 ++-
include/net/scm.h | 11 +
net/Makefile | 2 +-
net/core/scm.c | 17 ++
net/unix/Kconfig | 11 +-
net/unix/Makefile | 2 -
net/unix/af_unix.c | 120 +++++---
net/unix/garbage.c | 691 +++++++++++++++++++++++++++++-------------
net/unix/scm.c | 154 ----------
net/unix/scm.h | 10 -
10 files changed, 618 insertions(+), 448 deletions(-)
delete mode 100644 net/unix/scm.c
delete mode 100644 net/unix/scm.h
--
2.49.0.1143.g0be31eac6b-goog
Previously 1 is unconditionally taken as current brightness value. This
causes problems since it's required to restore brightness settings on
resumption, and a value that doesn't match EC's state before suspension
will cause surprising changes of screen brightness.
Let's get brightness from EC and take it as the current brightness on
probe of the laptop driver to avoid the surprising behavior. Tested on
TongFang L860-T2 3A5000 laptop.
Cc: stable(a)vger.kernel.org
Fixes: 6246ed09111f ("LoongArch: Add ACPI-based generic laptop driver")
Signed-off-by: Yao Zi <ziyao(a)disroot.org>
---
drivers/platform/loongarch/loongson-laptop.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/platform/loongarch/loongson-laptop.c b/drivers/platform/loongarch/loongson-laptop.c
index 99203584949d..828bd62e3596 100644
--- a/drivers/platform/loongarch/loongson-laptop.c
+++ b/drivers/platform/loongarch/loongson-laptop.c
@@ -392,7 +392,7 @@ static int laptop_backlight_register(void)
if (!acpi_evalf(hotkey_handle, &status, "ECLL", "d"))
return -EIO;
- props.brightness = 1;
+ props.brightness = ec_get_brightness();
props.max_brightness = status;
props.type = BACKLIGHT_PLATFORM;
--
2.49.0
Apply PCI host-bridge window offsets to screen_info framebuffers. Fixes
invalid access to I/O memory.
Resources behind a PCI host bridge can be relocated by a certain offset
in the kernel's CPU address range used for I/O. The framebuffer memory
range stored in screen_info refers to the CPU addresses as seen during
boot (where the offset is 0). During boot up, firmware may assign a
different memory offset to the PCI host bridge and thereby relocating
the framebuffer address of the PCI graphics device as seen by the kernel.
The information in screen_info must be updated as well.
The helper pcibios_bus_to_resource() performs the relocation of the
screen_info's framebuffer resource (given in PCI bus addresses). The
result matches the I/O-memory resource of the PCI graphics device (given
in CPU addresses). As before, we store away the information necessary to
later update the information in screen_info itself.
Commit 78aa89d1dfba ("firmware/sysfb: Update screen_info for relocated
EFI framebuffers") added the code for updating screen_info. It is based
on similar functionality that pre-existed in efifb. Efifb uses a pointer
to the PCI resource, while the newer code does a memcpy of the region.
Hence efifb sees any updates to the PCI resource and avoids the issue.
v3:
- Only use struct pci_bus_region for PCI bus addresses (Bjorn)
- Clarify address semantics in commit messages and comments (Bjorn)
v2:
- Fixed tags (Takashi, Ivan)
- Updated information on efifb
Signed-off-by: Thomas Zimmermann <tzimmermann(a)suse.de>
Reviewed-by: Javier Martinez Canillas <javierm(a)redhat.com>
Reported-by: "Ivan T. Ivanov" <iivanov(a)suse.de>
Closes: https://bugzilla.suse.com/show_bug.cgi?id=1240696
Tested-by: "Ivan T. Ivanov" <iivanov(a)suse.de>
Fixes: 78aa89d1dfba ("firmware/sysfb: Update screen_info for relocated EFI framebuffers")
Cc: dri-devel(a)lists.freedesktop.org
Cc: <stable(a)vger.kernel.org> # v6.9+
---
drivers/video/screen_info_pci.c | 79 +++++++++++++++++++++------------
1 file changed, 50 insertions(+), 29 deletions(-)
diff --git a/drivers/video/screen_info_pci.c b/drivers/video/screen_info_pci.c
index 6c5833517141..66bfc1d0a6dc 100644
--- a/drivers/video/screen_info_pci.c
+++ b/drivers/video/screen_info_pci.c
@@ -7,8 +7,8 @@
static struct pci_dev *screen_info_lfb_pdev;
static size_t screen_info_lfb_bar;
-static resource_size_t screen_info_lfb_offset;
-static struct resource screen_info_lfb_res = DEFINE_RES_MEM(0, 0);
+static resource_size_t screen_info_lfb_res_start; // original start of resource
+static resource_size_t screen_info_lfb_offset; // framebuffer offset within resource
static bool __screen_info_relocation_is_valid(const struct screen_info *si, struct resource *pr)
{
@@ -31,7 +31,7 @@ void screen_info_apply_fixups(void)
if (screen_info_lfb_pdev) {
struct resource *pr = &screen_info_lfb_pdev->resource[screen_info_lfb_bar];
- if (pr->start != screen_info_lfb_res.start) {
+ if (pr->start != screen_info_lfb_res_start) {
if (__screen_info_relocation_is_valid(si, pr)) {
/*
* Only update base if we have an actual
@@ -47,46 +47,67 @@ void screen_info_apply_fixups(void)
}
}
+static int __screen_info_lfb_pci_bus_region(const struct screen_info *si, unsigned int type,
+ struct pci_bus_region *r)
+{
+ u64 base, size;
+
+ base = __screen_info_lfb_base(si);
+ if (!base)
+ return -EINVAL;
+
+ size = __screen_info_lfb_size(si, type);
+ if (!size)
+ return -EINVAL;
+
+ r->start = base;
+ r->end = base + size - 1;
+
+ return 0;
+}
+
static void screen_info_fixup_lfb(struct pci_dev *pdev)
{
unsigned int type;
- struct resource res[SCREEN_INFO_MAX_RESOURCES];
- size_t i, numres;
+ struct pci_bus_region bus_region;
int ret;
+ struct resource r = {
+ .flags = IORESOURCE_MEM,
+ };
+ const struct resource *pr;
const struct screen_info *si = &screen_info;
if (screen_info_lfb_pdev)
return; // already found
type = screen_info_video_type(si);
- if (type != VIDEO_TYPE_EFI)
- return; // only applies to EFI
+ if (!__screen_info_has_lfb(type))
+ return; // only applies to EFI; maybe VESA
- ret = screen_info_resources(si, res, ARRAY_SIZE(res));
+ ret = __screen_info_lfb_pci_bus_region(si, type, &bus_region);
if (ret < 0)
return;
- numres = ret;
- for (i = 0; i < numres; ++i) {
- struct resource *r = &res[i];
- const struct resource *pr;
-
- if (!(r->flags & IORESOURCE_MEM))
- continue;
- pr = pci_find_resource(pdev, r);
- if (!pr)
- continue;
-
- /*
- * We've found a PCI device with the framebuffer
- * resource. Store away the parameters to track
- * relocation of the framebuffer aperture.
- */
- screen_info_lfb_pdev = pdev;
- screen_info_lfb_bar = pr - pdev->resource;
- screen_info_lfb_offset = r->start - pr->start;
- memcpy(&screen_info_lfb_res, r, sizeof(screen_info_lfb_res));
- }
+ /*
+ * Translate the PCI bus address to resource. Account
+ * for an offset if the framebuffer is behind a PCI host
+ * bridge.
+ */
+ pcibios_bus_to_resource(pdev->bus, &r, &bus_region);
+
+ pr = pci_find_resource(pdev, &r);
+ if (!pr)
+ return;
+
+ /*
+ * We've found a PCI device with the framebuffer
+ * resource. Store away the parameters to track
+ * relocation of the framebuffer aperture.
+ */
+ screen_info_lfb_pdev = pdev;
+ screen_info_lfb_bar = pr - pdev->resource;
+ screen_info_lfb_offset = r.start - pr->start;
+ screen_info_lfb_res_start = bus_region.start;
}
DECLARE_PCI_FIXUP_CLASS_HEADER(PCI_ANY_ID, PCI_ANY_ID, PCI_BASE_CLASS_DISPLAY, 16,
screen_info_fixup_lfb);
--
2.49.0
Object creation is a careful dance where we must guarantee that the
object is fully constructed before it is visible to other threads, and
GEM buffer objects are no difference.
Final publishing happens by calling drm_gem_handle_create(). After
that the only allowed thing to do is call drm_gem_object_put() because
a concurrent call to the GEM_CLOSE ioctl with a correctly guessed id
(which is trivial since we have a linear allocator) can already tear
down the object again.
Luckily most drivers get this right, the very few exceptions I've
pinged the relevant maintainers for. Unfortunately we also need
drm_gem_handle_create() when creating additional handles for an
already existing object (e.g. GETFB ioctl or the various bo import
ioctl), and hence we cannot have a drm_gem_handle_create_and_put() as
the only exported function to stop these issues from happening.
Now unfortunately the implementation of drm_gem_handle_create() isn't
living up to standards: It does correctly finishe object
initialization at the global level, and hence is safe against a
concurrent tear down. But it also sets up the file-private aspects of
the handle, and that part goes wrong: We fully register the object in
the drm_file.object_idr before calling drm_vma_node_allow() or
obj->funcs->open, which opens up races against concurrent removal of
that handle in drm_gem_handle_delete().
Fix this with the usual two-stage approach of first reserving the
handle id, and then only registering the object after we've completed
the file-private setup.
Jacek reported this with a testcase of concurrently calling GEM_CLOSE
on a freshly-created object (which also destroys the object), but it
should be possible to hit this with just additional handles created
through import or GETFB without completed destroying the underlying
object with the concurrent GEM_CLOSE ioctl calls.
Note that the close-side of this race was fixed in f6cd7daecff5 ("drm:
Release driver references to handle before making it available
again"), which means a cool 9 years have passed until someone noticed
that we need to make this symmetry or there's still gaps left :-/
Without the 2-stage close approach we'd still have a race, therefore
that's an integral part of this bugfix.
More importantly, this means we can have NULL pointers behind
allocated id in our drm_file.object_idr. We need to check for that
now:
- drm_gem_handle_delete() checks for ERR_OR_NULL already
- drm_gem.c:object_lookup() also chekcs for NULL
- drm_gem_release() should never be called if there's another thread
still existing that could call into an IOCTL that creates a new
handle, so cannot race. For paranoia I added a NULL check to
drm_gem_object_release_handle() though.
- most drivers (etnaviv, i915, msm) are find because they use
idr_find, which maps both ENOENT and NULL to NULL.
- vmgfx is already broken vmw_debugfs_gem_info_show() because NULL
pointers might exist due to drm_gem_handle_delete(). This needs a
separate patch. This is because idr_for_each_entry terminates on the
first NULL entry and so might not iterate over everything.
- similar for amd in amdgpu_debugfs_gem_info_show() and
amdgpu_gem_force_release(). The latter is really questionable though
since it's a best effort hack and there's no way to close all the
races. Needs separate patches.
- xe is really broken because it not uses idr_for_each_entry() but
also drops the drm_file.table_lock, which can wreak the idr iterator
state if you're unlucky enough. Maybe another reason to look into
the drm fdinfo memory stats instead of hand-rolling too much.
- drm_show_memory_stats() is also broken since it uses
idr_for_each_entry. But since that's a preexisting bug I'll follow
up with a separate patch.
Reported-by: Jacek Lawrynowicz <jacek.lawrynowicz(a)linux.intel.com>
Cc: stable(a)vger.kernel.org
Cc: Jacek Lawrynowicz <jacek.lawrynowicz(a)linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst(a)linux.intel.com>
Cc: Maxime Ripard <mripard(a)kernel.org>
Cc: Thomas Zimmermann <tzimmermann(a)suse.de>
Cc: David Airlie <airlied(a)gmail.com>
Cc: Simona Vetter <simona(a)ffwll.ch>
Signed-off-by: Simona Vetter <simona.vetter(a)intel.com>
Signed-off-by: Simona Vetter <simona.vetter(a)ffwll.ch>
---
drivers/gpu/drm/drm_gem.c | 10 +++++++++-
include/drm/drm_file.h | 3 +++
2 files changed, 12 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index 1e659d2660f7..e4e20dda47b1 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -279,6 +279,9 @@ drm_gem_object_release_handle(int id, void *ptr, void *data)
struct drm_file *file_priv = data;
struct drm_gem_object *obj = ptr;
+ if (WARN_ON(!data))
+ return 0;
+
if (obj->funcs->close)
obj->funcs->close(obj, file_priv);
@@ -399,7 +402,7 @@ drm_gem_handle_create_tail(struct drm_file *file_priv,
idr_preload(GFP_KERNEL);
spin_lock(&file_priv->table_lock);
- ret = idr_alloc(&file_priv->object_idr, obj, 1, 0, GFP_NOWAIT);
+ ret = idr_alloc(&file_priv->object_idr, NULL, 1, 0, GFP_NOWAIT);
spin_unlock(&file_priv->table_lock);
idr_preload_end();
@@ -420,6 +423,11 @@ drm_gem_handle_create_tail(struct drm_file *file_priv,
goto err_revoke;
}
+ /* mirrors drm_gem_handle_delete to avoid races */
+ spin_lock(&file_priv->table_lock);
+ obj = idr_replace(&file_priv->object_idr, obj, handle);
+ WARN_ON(obj != NULL);
+ spin_unlock(&file_priv->table_lock);
*handlep = handle;
return 0;
diff --git a/include/drm/drm_file.h b/include/drm/drm_file.h
index 5c3b2aa3e69d..d344d41e6cfe 100644
--- a/include/drm/drm_file.h
+++ b/include/drm/drm_file.h
@@ -300,6 +300,9 @@ struct drm_file {
*
* Mapping of mm object handles to object pointers. Used by the GEM
* subsystem. Protected by @table_lock.
+ *
+ * Note that allocated entries might be NULL as a transient state when
+ * creating or deleting a handle.
*/
struct idr object_idr;
--
2.49.0
Add the missing memory barriers to make sure that destination ring
descriptors are read after the head pointers to avoid using stale data
on weakly ordered architectures like aarch64.
Note that this may fix the empty descriptor issue recently worked around
by commit 51ad34a47e9f ("wifi: ath12k: Add drop descriptor handling for
monitor ring").
Tested-on: WCN7850 hw2.0 WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Fixes: d889913205cf ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Cc: stable(a)vger.kernel.org # 6.3
Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org>
---
drivers/net/wireless/ath/ath12k/dp_mon.c | 3 +++
drivers/net/wireless/ath/ath12k/dp_rx.c | 12 ++++++++++++
drivers/net/wireless/ath/ath12k/dp_tx.c | 3 +++
3 files changed, 18 insertions(+)
diff --git a/drivers/net/wireless/ath/ath12k/dp_mon.c b/drivers/net/wireless/ath/ath12k/dp_mon.c
index d22800e89485..90a7763502c8 100644
--- a/drivers/net/wireless/ath/ath12k/dp_mon.c
+++ b/drivers/net/wireless/ath/ath12k/dp_mon.c
@@ -3258,6 +3258,9 @@ int ath12k_dp_mon_srng_process(struct ath12k *ar, int *budget,
spin_lock_bh(&srng->lock);
ath12k_hal_srng_access_begin(ab, srng);
+ /* Make sure descriptor is read after the head pointer. */
+ dma_rmb();
+
while (likely(*budget)) {
*budget -= 1;
mon_dst_desc = ath12k_hal_srng_dst_peek(ab, srng);
diff --git a/drivers/net/wireless/ath/ath12k/dp_rx.c b/drivers/net/wireless/ath/ath12k/dp_rx.c
index 75bf4211ad42..68fceb4201d7 100644
--- a/drivers/net/wireless/ath/ath12k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath12k/dp_rx.c
@@ -2753,6 +2753,9 @@ int ath12k_dp_rx_process(struct ath12k_base *ab, int ring_id,
try_again:
ath12k_hal_srng_access_begin(ab, srng);
+ /* Make sure descriptor is read after the head pointer. */
+ dma_rmb();
+
while ((desc = ath12k_hal_srng_dst_get_next_entry(ab, srng))) {
struct rx_mpdu_desc *mpdu_info;
struct rx_msdu_desc *msdu_info;
@@ -3599,6 +3602,9 @@ int ath12k_dp_rx_process_err(struct ath12k_base *ab, struct napi_struct *napi,
ath12k_hal_srng_access_begin(ab, srng);
+ /* Make sure descriptor is read after the head pointer. */
+ dma_rmb();
+
while (budget &&
(reo_desc = ath12k_hal_srng_dst_get_next_entry(ab, srng))) {
drop = false;
@@ -3941,6 +3947,9 @@ int ath12k_dp_rx_process_wbm_err(struct ath12k_base *ab,
ath12k_hal_srng_access_begin(ab, srng);
+ /* Make sure descriptor is read after the head pointer. */
+ dma_rmb();
+
while (budget) {
rx_desc = ath12k_hal_srng_dst_get_next_entry(ab, srng);
if (!rx_desc)
@@ -4122,6 +4131,9 @@ void ath12k_dp_rx_process_reo_status(struct ath12k_base *ab)
ath12k_hal_srng_access_begin(ab, srng);
+ /* Make sure descriptor is read after the head pointer. */
+ dma_rmb();
+
while ((hdr = ath12k_hal_srng_dst_get_next_entry(ab, srng))) {
tag = le64_get_bits(hdr->tl, HAL_SRNG_TLV_HDR_TAG);
diff --git a/drivers/net/wireless/ath/ath12k/dp_tx.c b/drivers/net/wireless/ath/ath12k/dp_tx.c
index ced232bf4aed..3124eafa0201 100644
--- a/drivers/net/wireless/ath/ath12k/dp_tx.c
+++ b/drivers/net/wireless/ath/ath12k/dp_tx.c
@@ -853,6 +853,9 @@ void ath12k_dp_tx_completion_handler(struct ath12k_base *ab, int ring_id)
ath12k_hal_srng_access_begin(ab, status_ring);
+ /* Make sure descriptor is read after the head pointer. */
+ dma_rmb();
+
while (ATH12K_TX_COMPL_NEXT(tx_ring->tx_status_head) != tx_ring->tx_status_tail) {
desc = ath12k_hal_srng_dst_get_next_entry(ab, status_ring);
if (!desc)
--
2.49.0