- Linux-stable-mirror - lists.linaro.org

[PATCH 5.15.y] RDMA/core: Refactor rdma_bind_addr

by Zhi Yang

From: Patrisious Haddad <phaddad(a)nvidia.com> commit 8d037973d48c026224ab285e6a06985ccac6f7bf upstream. Refactor rdma_bind_addr function so that it doesn't require that the cma destination address be changed before calling it. So now it will update the destination address internally only when it is really needed and after passing all the required checks. Which in turn results in a cleaner and more sensible call and error handling flows for the functions that call it directly or indirectly. Signed-off-by: Patrisious Haddad <phaddad(a)nvidia.com> Reported-by: Wei Chen <harperchen1110(a)gmail.com> Reviewed-by: Mark Zhang <markzhang(a)nvidia.com> Link: https://lore.kernel.org/r/3d0e9a2fd62bc10ba02fed1c7c48a48638952320.16728192… Signed-off-by: Leon Romanovsky <leon(a)kernel.org> Signed-off-by: Zhi Yang <Zhi.Yang(a)windriver.com> Signed-off-by: He Zhe <zhe.he(a)windriver.com> --- Build test passed. --- drivers/infiniband/core/cma.c | 253 +++++++++++++++++----------------- 1 file changed, 130 insertions(+), 123 deletions(-) diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index a204c738e9cf..e8dfce17a5e5 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -3364,121 +3364,6 @@ static int cma_resolve_ib_addr(struct rdma_id_private *id_priv) return ret; } -static int cma_bind_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, - const struct sockaddr *dst_addr) -{ - struct sockaddr_storage zero_sock = {}; - - if (src_addr && src_addr->sa_family) - return rdma_bind_addr(id, src_addr); - - /* - * When the src_addr is not specified, automatically supply an any addr - */ - zero_sock.ss_family = dst_addr->sa_family; - if (IS_ENABLED(CONFIG_IPV6) && dst_addr->sa_family == AF_INET6) { - struct sockaddr_in6 *src_addr6 = - (struct sockaddr_in6 *)&zero_sock; - struct sockaddr_in6 *dst_addr6 = - (struct sockaddr_in6 *)dst_addr; - - src_addr6->sin6_scope_id = dst_addr6->sin6_scope_id; - if (ipv6_addr_type(&dst_addr6->sin6_addr) & IPV6_ADDR_LINKLOCAL) - id->route.addr.dev_addr.bound_dev_if = - dst_addr6->sin6_scope_id; - } else if (dst_addr->sa_family == AF_IB) { - ((struct sockaddr_ib *)&zero_sock)->sib_pkey = - ((struct sockaddr_ib *)dst_addr)->sib_pkey; - } - return rdma_bind_addr(id, (struct sockaddr *)&zero_sock); -} - -/* - * If required, resolve the source address for bind and leave the id_priv in - * state RDMA_CM_ADDR_BOUND. This oddly uses the state to determine the prior - * calls made by ULP, a previously bound ID will not be re-bound and src_addr is - * ignored. - */ -static int resolve_prepare_src(struct rdma_id_private *id_priv, - struct sockaddr *src_addr, - const struct sockaddr *dst_addr) -{ - int ret; - - memcpy(cma_dst_addr(id_priv), dst_addr, rdma_addr_size(dst_addr)); - if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_ADDR_QUERY)) { - /* For a well behaved ULP state will be RDMA_CM_IDLE */ - ret = cma_bind_addr(&id_priv->id, src_addr, dst_addr); - if (ret) - goto err_dst; - if (WARN_ON(!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, - RDMA_CM_ADDR_QUERY))) { - ret = -EINVAL; - goto err_dst; - } - } - - if (cma_family(id_priv) != dst_addr->sa_family) { - ret = -EINVAL; - goto err_state; - } - return 0; - -err_state: - cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); -err_dst: - memset(cma_dst_addr(id_priv), 0, rdma_addr_size(dst_addr)); - return ret; -} - -int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, - const struct sockaddr *dst_addr, unsigned long timeout_ms) -{ - struct rdma_id_private *id_priv = - container_of(id, struct rdma_id_private, id); - int ret; - - ret = resolve_prepare_src(id_priv, src_addr, dst_addr); - if (ret) - return ret; - - if (cma_any_addr(dst_addr)) { - ret = cma_resolve_loopback(id_priv); - } else { - if (dst_addr->sa_family == AF_IB) { - ret = cma_resolve_ib_addr(id_priv); - } else { - /* - * The FSM can return back to RDMA_CM_ADDR_BOUND after - * rdma_resolve_ip() is called, eg through the error - * path in addr_handler(). If this happens the existing - * request must be canceled before issuing a new one. - * Since canceling a request is a bit slow and this - * oddball path is rare, keep track once a request has - * been issued. The track turns out to be a permanent - * state since this is the only cancel as it is - * immediately before rdma_resolve_ip(). - */ - if (id_priv->used_resolve_ip) - rdma_addr_cancel(&id->route.addr.dev_addr); - else - id_priv->used_resolve_ip = 1; - ret = rdma_resolve_ip(cma_src_addr(id_priv), dst_addr, - &id->route.addr.dev_addr, - timeout_ms, addr_handler, - false, id_priv); - } - } - if (ret) - goto err; - - return 0; -err: - cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); - return ret; -} -EXPORT_SYMBOL(rdma_resolve_addr); - int rdma_set_reuseaddr(struct rdma_cm_id *id, int reuse) { struct rdma_id_private *id_priv; @@ -3881,27 +3766,26 @@ int rdma_listen(struct rdma_cm_id *id, int backlog) } EXPORT_SYMBOL(rdma_listen); -int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) +static int rdma_bind_addr_dst(struct rdma_id_private *id_priv, + struct sockaddr *addr, const struct sockaddr *daddr) { - struct rdma_id_private *id_priv; + struct sockaddr *id_daddr; int ret; - struct sockaddr *daddr; if (addr->sa_family != AF_INET && addr->sa_family != AF_INET6 && addr->sa_family != AF_IB) return -EAFNOSUPPORT; - id_priv = container_of(id, struct rdma_id_private, id); if (!cma_comp_exch(id_priv, RDMA_CM_IDLE, RDMA_CM_ADDR_BOUND)) return -EINVAL; - ret = cma_check_linklocal(&id->route.addr.dev_addr, addr); + ret = cma_check_linklocal(&id_priv->id.route.addr.dev_addr, addr); if (ret) goto err1; memcpy(cma_src_addr(id_priv), addr, rdma_addr_size(addr)); if (!cma_any_addr(addr)) { - ret = cma_translate_addr(addr, &id->route.addr.dev_addr); + ret = cma_translate_addr(addr, &id_priv->id.route.addr.dev_addr); if (ret) goto err1; @@ -3921,8 +3805,10 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) } #endif } - daddr = cma_dst_addr(id_priv); - daddr->sa_family = addr->sa_family; + id_daddr = cma_dst_addr(id_priv); + if (daddr != id_daddr) + memcpy(id_daddr, daddr, rdma_addr_size(addr)); + id_daddr->sa_family = addr->sa_family; ret = cma_get_port(id_priv); if (ret) @@ -3938,6 +3824,127 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE); return ret; } + +static int cma_bind_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, + const struct sockaddr *dst_addr) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + struct sockaddr_storage zero_sock = {}; + + if (src_addr && src_addr->sa_family) + return rdma_bind_addr_dst(id_priv, src_addr, dst_addr); + + /* + * When the src_addr is not specified, automatically supply an any addr + */ + zero_sock.ss_family = dst_addr->sa_family; + if (IS_ENABLED(CONFIG_IPV6) && dst_addr->sa_family == AF_INET6) { + struct sockaddr_in6 *src_addr6 = + (struct sockaddr_in6 *)&zero_sock; + struct sockaddr_in6 *dst_addr6 = + (struct sockaddr_in6 *)dst_addr; + + src_addr6->sin6_scope_id = dst_addr6->sin6_scope_id; + if (ipv6_addr_type(&dst_addr6->sin6_addr) & IPV6_ADDR_LINKLOCAL) + id->route.addr.dev_addr.bound_dev_if = + dst_addr6->sin6_scope_id; + } else if (dst_addr->sa_family == AF_IB) { + ((struct sockaddr_ib *)&zero_sock)->sib_pkey = + ((struct sockaddr_ib *)dst_addr)->sib_pkey; + } + return rdma_bind_addr_dst(id_priv, (struct sockaddr *)&zero_sock, dst_addr); +} + +/* + * If required, resolve the source address for bind and leave the id_priv in + * state RDMA_CM_ADDR_BOUND. This oddly uses the state to determine the prior + * calls made by ULP, a previously bound ID will not be re-bound and src_addr is + * ignored. + */ +static int resolve_prepare_src(struct rdma_id_private *id_priv, + struct sockaddr *src_addr, + const struct sockaddr *dst_addr) +{ + int ret; + + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_ADDR_QUERY)) { + /* For a well behaved ULP state will be RDMA_CM_IDLE */ + ret = cma_bind_addr(&id_priv->id, src_addr, dst_addr); + if (ret) + return ret; + if (WARN_ON(!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, + RDMA_CM_ADDR_QUERY))) + return -EINVAL; + + } + + if (cma_family(id_priv) != dst_addr->sa_family) { + ret = -EINVAL; + goto err_state; + } + return 0; + +err_state: + cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); + return ret; +} + +int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, + const struct sockaddr *dst_addr, unsigned long timeout_ms) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + int ret; + + ret = resolve_prepare_src(id_priv, src_addr, dst_addr); + if (ret) + return ret; + + if (cma_any_addr(dst_addr)) { + ret = cma_resolve_loopback(id_priv); + } else { + if (dst_addr->sa_family == AF_IB) { + ret = cma_resolve_ib_addr(id_priv); + } else { + /* + * The FSM can return back to RDMA_CM_ADDR_BOUND after + * rdma_resolve_ip() is called, eg through the error + * path in addr_handler(). If this happens the existing + * request must be canceled before issuing a new one. + * Since canceling a request is a bit slow and this + * oddball path is rare, keep track once a request has + * been issued. The track turns out to be a permanent + * state since this is the only cancel as it is + * immediately before rdma_resolve_ip(). + */ + if (id_priv->used_resolve_ip) + rdma_addr_cancel(&id->route.addr.dev_addr); + else + id_priv->used_resolve_ip = 1; + ret = rdma_resolve_ip(cma_src_addr(id_priv), dst_addr, + &id->route.addr.dev_addr, + timeout_ms, addr_handler, + false, id_priv); + } + } + if (ret) + goto err; + + return 0; +err: + cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); + return ret; +} +EXPORT_SYMBOL(rdma_resolve_addr); + +int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + + return rdma_bind_addr_dst(id_priv, addr, cma_dst_addr(id_priv)); +} EXPORT_SYMBOL(rdma_bind_addr); static int cma_format_hdr(void *hdr, struct rdma_id_private *id_priv) -- 2.34.1

1 day, 11 hours

1
0
0 0

[PATCH net] net: qede: Initialize qede_ll_ops with designated initializer

by Nathan Chancellor

After a recent change [1] in clang's randstruct implementation to randomize structures that only contain function pointers, there is an error because qede_ll_ops get randomized but does not use a designated initializer for the first member: drivers/net/ethernet/qlogic/qede/qede_main.c:206:2: error: a randomized struct can only be initialized with a designated initializer 206 | { | ^ Explicitly initialize the common member using a designated initializer to fix the build. Cc: stable(a)vger.kernel.org Fixes: 035f7f87b729 ("randstruct: Enable Clang support") Link: https://github.com/llvm/llvm-project/commit/04364fb888eea6db9811510607bed4b… [1] Signed-off-by: Nathan Chancellor <nathan(a)kernel.org> --- drivers/net/ethernet/qlogic/qede/qede_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/net/ethernet/qlogic/qede/qede_main.c b/drivers/net/ethernet/qlogic/qede/qede_main.c index 99df00c30b8c..b5d744d2586f 100644 --- a/drivers/net/ethernet/qlogic/qede/qede_main.c +++ b/drivers/net/ethernet/qlogic/qede/qede_main.c @@ -203,7 +203,7 @@ static struct pci_driver qede_pci_driver = { }; static struct qed_eth_cb_ops qede_ll_ops = { - { + .common = { #ifdef CONFIG_RFS_ACCEL .arfs_filter_op = qede_arfs_filter_op, #endif --- base-commit: 9540984da649d46f699c47f28c68bbd3c9d99e4c change-id: 20250507-qede-fix-clang-randstruct-13d8c593cb58 Best regards, -- Nathan Chancellor <nathan(a)kernel.org>

1 day, 11 hours

3
2
0 0

[PATCH 5.10.y] drm/vmwgfx: Fix a deadlock in dma buf fence polling

by Zhi Yang

From: Zack Rusin <zack.rusin(a)broadcom.com> commit e58337100721f3cc0c7424a18730e4f39844934f upstream. Introduce a version of the fence ops that on release doesn't remove the fence from the pending list, and thus doesn't require a lock to fix poll->fence wait->fence unref deadlocks. vmwgfx overwrites the wait callback to iterate over the list of all fences and update their status, to do that it holds a lock to prevent the list modifcations from other threads. The fence destroy callback both deletes the fence and removes it from the list of pending fences, for which it holds a lock. dma buf polling cb unrefs a fence after it's been signaled: so the poll calls the wait, which signals the fences, which are being destroyed. The destruction tries to acquire the lock on the pending fences list which it can never get because it's held by the wait from which it was called. Old bug, but not a lot of userspace apps were using dma-buf polling interfaces. Fix those, in particular this fixes KDE stalls/deadlock. Signed-off-by: Zack Rusin <zack.rusin(a)broadcom.com> Fixes: 2298e804e96e ("drm/vmwgfx: rework to new fence interface, v2") Cc: Broadcom internal kernel review list <bcm-kernel-feedback-list(a)broadcom.com> Cc: dri-devel(a)lists.freedesktop.org Cc: <stable(a)vger.kernel.org> # v6.2+ Reviewed-by: Maaz Mombasawala <maaz.mombasawala(a)broadcom.com> Reviewed-by: Martin Krastev <martin.krastev(a)broadcom.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240722184313.181318-2-zack.… [Minor context change fixed] Signed-off-by: Zhi Yang <Zhi.Yang(a)windriver.com> Signed-off-by: He Zhe <zhe.he(a)windriver.com> --- Build test passed. --- drivers/gpu/drm/vmwgfx/vmwgfx_fence.c | 17 +++++++---------- 1 file changed, 7 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c index 6bacdb7583df..0505f87d13c0 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c @@ -32,7 +32,6 @@ #define VMW_FENCE_WRAP (1 << 31) struct vmw_fence_manager { - int num_fence_objects; struct vmw_private *dev_priv; spinlock_t lock; struct list_head fence_list; @@ -113,13 +112,13 @@ static void vmw_fence_obj_destroy(struct dma_fence *f) { struct vmw_fence_obj *fence = container_of(f, struct vmw_fence_obj, base); - struct vmw_fence_manager *fman = fman_from_fence(fence); - spin_lock(&fman->lock); - list_del_init(&fence->head); - --fman->num_fence_objects; - spin_unlock(&fman->lock); + if (!list_empty(&fence->head)) { + spin_lock(&fman->lock); + list_del_init(&fence->head); + spin_unlock(&fman->lock); + } fence->destroy(fence); } @@ -250,7 +249,6 @@ static const struct dma_fence_ops vmw_fence_ops = { .release = vmw_fence_obj_destroy, }; - /** * Execute signal actions on fences recently signaled. * This is done from a workqueue so we don't have to execute @@ -353,7 +351,6 @@ static int vmw_fence_obj_init(struct vmw_fence_manager *fman, goto out_unlock; } list_add_tail(&fence->head, &fman->fence_list); - ++fman->num_fence_objects; out_unlock: spin_unlock(&fman->lock); @@ -402,7 +399,7 @@ static bool vmw_fence_goal_new_locked(struct vmw_fence_manager *fman, { u32 goal_seqno; u32 *fifo_mem; - struct vmw_fence_obj *fence; + struct vmw_fence_obj *fence, *next_fence; if (likely(!fman->seqno_valid)) return false; @@ -413,7 +410,7 @@ static bool vmw_fence_goal_new_locked(struct vmw_fence_manager *fman, return false; fman->seqno_valid = false; - list_for_each_entry(fence, &fman->fence_list, head) { + list_for_each_entry_safe(fence, next_fence, &fman->fence_list, head) { if (!list_empty(&fence->seq_passed_actions)) { fman->seqno_valid = true; vmw_mmio_write(fence->base.seqno, -- 2.34.1

1 day, 12 hours

1
0
0 0

[PATCH] LoongArch: Prevent cond_resched() occurring within kernel-fpu

by Tianyang Zhang

When CONFIG_PREEMPT_COUNT is not configured (i.e. CONFIG_PREEMPT_NONE/ CONFIG_PREEMPT_VOLUNTARY), preempt_disable() / preempt_enable() merely acts as a barrier(). However, in these cases cond_resched() can still trigger a context switch and modify the CSR.EUEN, resulting in do_fpu() exception being activated within the kernel-fpu critical sections, as demonstrated in the following path: dcn32_calculate_wm_and_dlg() DC_FP_START() dcn32_calculate_wm_and_dlg_fpu() dcn32_find_dummy_latency_index_for_fw_based_mclk_switch() dcn32_internal_validate_bw() dcn32_enable_phantom_stream() dc_create_stream_for_sink() kzalloc(GFP_KERNEL) __kmem_cache_alloc_node() __cond_resched() DC_FP_END() This patch is similar to commit d021985 (x86/fpu: Improve crypto performance by making kernel-mode FPU reliably usable in softirqs). It uses local_bh_disable() instead of preempt_disable() for non-RT kernels so it can avoid the cond_resched() issue, and also extend the kernel-fpu application scenarios to the softirq context. Cc: stable(a)vger.kernel.org Signed-off-by: Tianyang Zhang <zhangtianyang(a)loongson.cn> Signed-off-by: Huacai Chen <chenhuacai(a)loongson.cn> --- arch/loongarch/kernel/kfpu.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) diff --git a/arch/loongarch/kernel/kfpu.c b/arch/loongarch/kernel/kfpu.c index ec5b28e570c9..4e469b021cf4 100644 --- a/arch/loongarch/kernel/kfpu.c +++ b/arch/loongarch/kernel/kfpu.c @@ -18,11 +18,28 @@ static unsigned int euen_mask = CSR_EUEN_FPEN; static DEFINE_PER_CPU(bool, in_kernel_fpu); static DEFINE_PER_CPU(unsigned int, euen_current); +static inline void fpregs_lock(void) +{ + if (!IS_ENABLED(CONFIG_PREEMPT_RT)) + local_bh_disable(); + else + preempt_disable(); +} + +static inline void fpregs_unlock(void) +{ + if (!IS_ENABLED(CONFIG_PREEMPT_RT)) + local_bh_enable(); + else + preempt_enable(); +} + void kernel_fpu_begin(void) { unsigned int *euen_curr; - preempt_disable(); + if (!irqs_disabled()) + fpregs_lock(); WARN_ON(this_cpu_read(in_kernel_fpu)); @@ -73,7 +90,8 @@ void kernel_fpu_end(void) this_cpu_write(in_kernel_fpu, false); - preempt_enable(); + if (!irqs_disabled()) + fpregs_unlock(); } EXPORT_SYMBOL_GPL(kernel_fpu_end); -- 2.20.1

1 day, 12 hours

1
0
0 0

Re: [PATCH] um: work around sched_yield not yielding in time-travel mode

by Sasha Levin

[ Sasha's backport helper bot ] Hi, Summary of potential issues: ⚠️ Found matching upstream commit but patch is missing proper reference to it Found matching upstream commit: 887c5c12e80c8424bd471122d2e8b6b462e12874 WARNING: Author mismatch between patch and found commit: Backport author: Benjamin Berg<benjamin(a)sipsolutions.net> Commit author: Benjamin Berg<benjamin.berg(a)intel.com> Note: The patch differs from the upstream commit: --- 1: 887c5c12e80c8 < -: ------------- um: work around sched_yield not yielding in time-travel mode -: ------------- > 1: aeaee199900ee Linux 6.14.5 --- Results of testing on various branches: | Branch | Patch Apply | Build Test | |---------------------------|-------------|------------| | stable/linux-6.14.y | Success | Success | | stable/linux-6.12.y | Success | Success | | stable/linux-6.6.y | Success | Success | | stable/linux-6.1.y | Success | Success | | stable/linux-5.15.y | Success | Success | | stable/linux-5.10.y | Success | Success | | stable/linux-5.4.y | Success | Success |

1 day, 12 hours

1
0
0 0

[PATCH 6.6.y] btrfs: always fallback to buffered write if the inode requires checksum

by Qu Wenruo

commit 968f19c5b1b7d5595423b0ac0020cc18dfed8cb5 upstream. [BUG] It is a long known bug that VM image on btrfs can lead to data csum mismatch, if the qemu is using direct-io for the image (this is commonly known as cache mode 'none'). [CAUSE] Inside the VM, if the fs is EXT4 or XFS, or even NTFS from Windows, the fs is allowed to dirty/modify the folio even if the folio is under writeback (as long as the address space doesn't have AS_STABLE_WRITES flag inherited from the block device). This is a valid optimization to improve the concurrency, and since these filesystems have no extra checksum on data, the content change is not a problem at all. But the final write into the image file is handled by btrfs, which needs the content not to be modified during writeback, or the checksum will not match the data (checksum is calculated before submitting the bio). So EXT4/XFS/NTRFS assume they can modify the folio under writeback, but btrfs requires no modification, this leads to the false csum mismatch. This is only a controlled example, there are even cases where multi-thread programs can submit a direct IO write, then another thread modifies the direct IO buffer for whatever reason. For such cases, btrfs has no sane way to detect such cases and leads to false data csum mismatch. [FIX] I have considered the following ideas to solve the problem: - Make direct IO to always skip data checksum This not only requires a new incompatible flag, as it breaks the current per-inode NODATASUM flag. But also requires extra handling for no csum found cases. And this also reduces our checksum protection. - Let hardware handle all the checksum AKA, just nodatasum mount option. That requires trust for hardware (which is not that trustful in a lot of cases), and it's not generic at all. - Always fallback to buffered write if the inode requires checksum This was suggested by Christoph, and is the solution utilized by this patch. The cost is obvious, the extra buffer copying into page cache, thus it reduces the performance. But at least it's still user configurable, if the end user still wants the zero-copy performance, just set NODATASUM flag for the inode (which is a common practice for VM images on btrfs). Since we cannot trust user space programs to keep the buffer consistent during direct IO, we have no choice but always falling back to buffered IO. At least by this, we avoid the more deadly false data checksum mismatch error. CC: stable(a)vger.kernel.org # 6.6 Suggested-by: Christoph Hellwig <hch(a)infradead.org> Reviewed-by: Filipe Manana <fdmanana(a)suse.com> Signed-off-by: Qu Wenruo <wqu(a)suse.com> Reviewed-by: David Sterba <dsterba(a)suse.com> Signed-off-by: David Sterba <dsterba(a)suse.com> [ Fix a conflict due to the movement of the function. ] --- fs/btrfs/file.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index e794606e7c78..f1456c745c6d 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1515,6 +1515,23 @@ static ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from) goto buffered; } + /* + * We can't control the folios being passed in, applications can write + * to them while a direct IO write is in progress. This means the + * content might change after we calculated the data checksum. + * Therefore we can end up storing a checksum that doesn't match the + * persisted data. + * + * To be extra safe and avoid false data checksum mismatch, if the + * inode requires data checksum, just fallback to buffered IO. + * For buffered IO we have full control of page cache and can ensure + * no one is modifying the content during writeback. + */ + if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) { + btrfs_inode_unlock(BTRFS_I(inode), ilock_flags); + goto buffered; + } + /* * The iov_iter can be mapped to the same file range we are writing to. * If that's the case, then we will deadlock in the iomap code, because -- 2.49.0

1 day, 12 hours

4
5
0 0

[PATCH] bcachefs: Change btree_insert_node() assertion to error

by Kent Overstreet

Debug for https://github.com/koverstreet/bcachefs/issues/843 Print useful debug info and go emergency read-only. (cherry picked from commit 63c3b8f616cc95bb1fcc6101c92485d41c535d7c) Signed-off-by: Kent Overstreet <kent.overstreet(a)linux.dev> --- fs/bcachefs/btree_update_interior.c | 17 ++++++++++++++++- fs/bcachefs/error.c | 8 ++++++++ fs/bcachefs/error.h | 2 ++ 3 files changed, 26 insertions(+), 1 deletion(-) diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index e4e7c804625e..e9be8b5571a4 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -35,6 +35,8 @@ static const char * const bch2_btree_update_modes[] = { NULL }; +static void bch2_btree_update_to_text(struct printbuf *, struct btree_update *); + static int bch2_btree_insert_node(struct btree_update *, struct btree_trans *, btree_path_idx_t, struct btree *, struct keylist *); static void bch2_btree_update_add_new_node(struct btree_update *, struct btree *); @@ -1782,11 +1784,24 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t int ret; lockdep_assert_held(&c->gc_lock); - BUG_ON(!btree_node_intent_locked(path, b->c.level)); BUG_ON(!b->c.level); BUG_ON(!as || as->b); bch2_verify_keylist_sorted(keys); + if (!btree_node_intent_locked(path, b->c.level)) { + struct printbuf buf = PRINTBUF; + bch2_log_msg_start(c, &buf); + prt_printf(&buf, "%s(): node not locked at level %u\n", + __func__, b->c.level); + bch2_btree_update_to_text(&buf, as); + bch2_btree_path_to_text(&buf, trans, path_idx); + + bch2_print_string_as_lines(KERN_ERR, buf.buf); + printbuf_exit(&buf); + bch2_fs_emergency_read_only(c); + return -EIO; + } + ret = bch2_btree_node_lock_write(trans, path, &b->c); if (ret) return ret; diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c index 038da6a61f6b..6cbf4819e923 100644 --- a/fs/bcachefs/error.c +++ b/fs/bcachefs/error.c @@ -11,6 +11,14 @@ #define FSCK_ERR_RATELIMIT_NR 10 +void bch2_log_msg_start(struct bch_fs *c, struct printbuf *out) +{ +#ifdef BCACHEFS_LOG_PREFIX + prt_printf(out, bch2_log_msg(c, "")); +#endif + printbuf_indent_add(out, 2); +} + bool bch2_inconsistent_error(struct bch_fs *c) { set_bit(BCH_FS_error, &c->flags); diff --git a/fs/bcachefs/error.h b/fs/bcachefs/error.h index 7acf2a27ca28..5730eb6b2f38 100644 --- a/fs/bcachefs/error.h +++ b/fs/bcachefs/error.h @@ -18,6 +18,8 @@ struct work_struct; /* Error messages: */ +void bch2_log_msg_start(struct bch_fs *, struct printbuf *); + /* * Inconsistency errors: The on disk data is inconsistent. If these occur during * initial recovery, they don't indicate a bug in the running code - we walk all -- 2.49.0

1 day, 12 hours

2
1
0 0

[PATCH 5.10.y] of: module: add buffer overflow check in of_modalias()

by Uwe Kleine-König

From: Sergey Shtylyov <s.shtylyov(a)omp.ru> commit cf7385cb26ac4f0ee6c7385960525ad534323252 upstream. In of_modalias(), if the buffer happens to be too small even for the 1st snprintf() call, the len parameter will become negative and str parameter (if not NULL initially) will point beyond the buffer's end. Add the buffer overflow check after the 1st snprintf() call and fix such check after the strlen() call (accounting for the terminating NUL char). Fixes: bc575064d688 ("of/device: use of_property_for_each_string to parse compatible strings") Signed-off-by: Sergey Shtylyov <s.shtylyov(a)omp.ru> Link: https://lore.kernel.org/r/bbfc6be0-c687-62b6-d015-5141b93f313e@omp.ru Signed-off-by: Rob Herring <robh(a)kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> Signed-off-by: Uwe Kleine-König <ukleinek(a)debian.org> --- drivers/of/device.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/of/device.c b/drivers/of/device.c index 3a547793135c..93f08f18f6b3 100644 --- a/drivers/of/device.c +++ b/drivers/of/device.c @@ -231,14 +231,15 @@ static ssize_t of_device_get_modalias(struct device *dev, char *str, ssize_t len csize = snprintf(str, len, "of:N%pOFn%c%s", dev->of_node, 'T', of_node_get_device_type(dev->of_node)); tsize = csize; + if (csize >= len) + csize = len > 0 ? len - 1 : 0; len -= csize; - if (str) - str += csize; + str += csize; of_property_for_each_string(dev->of_node, "compatible", p, compat) { csize = strlen(compat) + 1; tsize += csize; - if (csize > len) + if (csize >= len) continue; csize = snprintf(str, len, "C%s", compat); base-commit: 024a4a45fdf87218e3c0925475b05a27bcea103f -- 2.47.2

1 day, 12 hours

2
1
0 0

[PATCH 5.15.y] of: module: add buffer overflow check in of_modalias()

by Uwe Kleine-König

From: Sergey Shtylyov <s.shtylyov(a)omp.ru> commit cf7385cb26ac4f0ee6c7385960525ad534323252 upstream. In of_modalias(), if the buffer happens to be too small even for the 1st snprintf() call, the len parameter will become negative and str parameter (if not NULL initially) will point beyond the buffer's end. Add the buffer overflow check after the 1st snprintf() call and fix such check after the strlen() call (accounting for the terminating NUL char). Fixes: bc575064d688 ("of/device: use of_property_for_each_string to parse compatible strings") Signed-off-by: Sergey Shtylyov <s.shtylyov(a)omp.ru> Link: https://lore.kernel.org/r/bbfc6be0-c687-62b6-d015-5141b93f313e@omp.ru Signed-off-by: Rob Herring <robh(a)kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> Signed-off-by: Uwe Kleine-König <ukleinek(a)debian.org> --- drivers/of/device.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/of/device.c b/drivers/of/device.c index 19c42a9dcba9..f503bb10b10b 100644 --- a/drivers/of/device.c +++ b/drivers/of/device.c @@ -257,14 +257,15 @@ static ssize_t of_device_get_modalias(struct device *dev, char *str, ssize_t len csize = snprintf(str, len, "of:N%pOFn%c%s", dev->of_node, 'T', of_node_get_device_type(dev->of_node)); tsize = csize; + if (csize >= len) + csize = len > 0 ? len - 1 : 0; len -= csize; - if (str) - str += csize; + str += csize; of_property_for_each_string(dev->of_node, "compatible", p, compat) { csize = strlen(compat) + 1; tsize += csize; - if (csize > len) + if (csize >= len) continue; csize = snprintf(str, len, "C%s", compat); base-commit: 16fdf2c7111bd6927f16c3e811f5086fecebbf00 -- 2.47.2

1 day, 12 hours

2
1
0 0

[PATCH 6.12.y] btrfs: always fallback to buffered write if the inode requires checksum

by Qu Wenruo

commit 968f19c5b1b7d5595423b0ac0020cc18dfed8cb5 upstream. [BUG] It is a long known bug that VM image on btrfs can lead to data csum mismatch, if the qemu is using direct-io for the image (this is commonly known as cache mode 'none'). [CAUSE] Inside the VM, if the fs is EXT4 or XFS, or even NTFS from Windows, the fs is allowed to dirty/modify the folio even if the folio is under writeback (as long as the address space doesn't have AS_STABLE_WRITES flag inherited from the block device). This is a valid optimization to improve the concurrency, and since these filesystems have no extra checksum on data, the content change is not a problem at all. Bu the final write into the image file is handled by btrfs, which needs the content not to be modified during writeback, or the checksum will not match the data (checksum is calculated before submitting the bio). So EXT4/XFS/NTRFS assume they can modify the folio under writeback, but btrfs requires no modification, this leads to the false csum mismatch. This is only a controlled example, there are even cases where multi-thread programs can submit a direct IO write, then another thread modifies the direct IO buffer for whatever reason. For such cases, btrfs has no sane way to detect such cases and leads to false data csum mismatch. [FIX] I have considered the following ideas to solve the problem: - Make direct IO to always skip data checksum This not only requires a new incompatible flag, as it breaks the current per-inode NODATASUM flag. But also requires extra handling for no csum found cases. And this also reduces our checksum protection. - Let hardware handle all the checksum AKA, just nodatasum mount option. That requires trust for hardware (which is not that trustful in a lot of cases), and it's not generic at all. - Always fallback to buffered write if the inode requires checksum This was suggested by Christoph, and is the solution utilized by this patch. The cost is obvious, the extra buffer copying into page cache, thus it reduces the performance. But at least it's still user configurable, if the end user still wants the zero-copy performance, just set NODATASUM flag for the inode (which is a common practice for VM images on btrfs). Since we cannot trust user space programs to keep the buffer consistent during direct IO, we have no choice but always falling back to buffered IO. At least by this, we avoid the more deadly false data checksum mismatch error. CC: stable(a)vger.kernel.org # 6.12+ Suggested-by: Christoph Hellwig <hch(a)infradead.org> Reviewed-by: Filipe Manana <fdmanana(a)suse.com> Signed-off-by: Qu Wenruo <wqu(a)suse.com> Reviewed-by: David Sterba <dsterba(a)suse.com> Signed-off-by: David Sterba <dsterba(a)suse.com> --- fs/btrfs/direct-io.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/fs/btrfs/direct-io.c b/fs/btrfs/direct-io.c index 8567af46e16f..eacbb74bf6bc 100644 --- a/fs/btrfs/direct-io.c +++ b/fs/btrfs/direct-io.c @@ -855,6 +855,22 @@ ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from) btrfs_inode_unlock(BTRFS_I(inode), ilock_flags); goto buffered; } + /* + * We can't control the folios being passed in, applications can write + * to them while a direct IO write is in progress. This means the + * content might change after we calculated the data checksum. + * Therefore we can end up storing a checksum that doesn't match the + * persisted data. + * + * To be extra safe and avoid false data checksum mismatch, if the + * inode requires data checksum, just fallback to buffered IO. + * For buffered IO we have full control of page cache and can ensure + * no one is modifying the content during writeback. + */ + if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) { + btrfs_inode_unlock(BTRFS_I(inode), ilock_flags); + goto buffered; + } /* * The iov_iter can be mapped to the same file range we are writing to. -- 2.49.0

1 day, 12 hours

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror