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

[PATCH v2] drm/xe: Defer buffer object shrinker write-backs and GPU waits

by Thomas Hellström

When the xe buffer-object shrinker allows GPU waits and write-back, (typically from kswapd), perform multiple passes, skipping subsequent passes if the shrinker number of scanned objects target is reached. 1) Without GPU waits and write-back 2) Without write-back 3) With both GPU-waits and write-back This is to avoid stalls and costly write- and readbacks unless they are really necessary. v2: - Don't test for scan completion twice. (Stuart Summers) - Update tags. Reported-by: melvyn <melvyn2(a)dnsense.pub> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/5557 Cc: Summers Stuart <stuart.summers(a)intel.com> Fixes: 00c8efc3180f ("drm/xe: Add a shrinker for xe bos") Cc: <stable(a)vger.kernel.org> # v6.15+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_shrinker.c | 51 +++++++++++++++++++++++++++++--- 1 file changed, 47 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_shrinker.c b/drivers/gpu/drm/xe/xe_shrinker.c index 1c3c04d52f55..90244fe59b59 100644 --- a/drivers/gpu/drm/xe/xe_shrinker.c +++ b/drivers/gpu/drm/xe/xe_shrinker.c @@ -54,10 +54,10 @@ xe_shrinker_mod_pages(struct xe_shrinker *shrinker, long shrinkable, long purgea write_unlock(&shrinker->lock); } -static s64 xe_shrinker_walk(struct xe_device *xe, - struct ttm_operation_ctx *ctx, - const struct xe_bo_shrink_flags flags, - unsigned long to_scan, unsigned long *scanned) +static s64 __xe_shrinker_walk(struct xe_device *xe, + struct ttm_operation_ctx *ctx, + const struct xe_bo_shrink_flags flags, + unsigned long to_scan, unsigned long *scanned) { unsigned int mem_type; s64 freed = 0, lret; @@ -93,6 +93,48 @@ static s64 xe_shrinker_walk(struct xe_device *xe, return freed; } +/* + * Try shrinking idle objects without writeback first, then if not sufficient, + * try also non-idle objects and finally if that's not sufficient either, + * add writeback. This avoids stalls and explicit writebacks with light or + * moderate memory pressure. + */ +static s64 xe_shrinker_walk(struct xe_device *xe, + struct ttm_operation_ctx *ctx, + const struct xe_bo_shrink_flags flags, + unsigned long to_scan, unsigned long *scanned) +{ + bool no_wait_gpu = true; + struct xe_bo_shrink_flags save_flags = flags; + s64 lret, freed; + + swap(no_wait_gpu, ctx->no_wait_gpu); + save_flags.writeback = false; + lret = __xe_shrinker_walk(xe, ctx, save_flags, to_scan, scanned); + swap(no_wait_gpu, ctx->no_wait_gpu); + if (lret < 0 || *scanned >= to_scan) + return lret; + + freed = lret; + if (!ctx->no_wait_gpu) { + lret = __xe_shrinker_walk(xe, ctx, save_flags, to_scan, scanned); + if (lret < 0) + return lret; + freed += lret; + if (*scanned >= to_scan) + return freed; + } + + if (flags.writeback) { + lret = __xe_shrinker_walk(xe, ctx, flags, to_scan, scanned); + if (lret < 0) + return lret; + freed += lret; + } + + return freed; +} + static unsigned long xe_shrinker_count(struct shrinker *shrink, struct shrink_control *sc) { @@ -199,6 +241,7 @@ static unsigned long xe_shrinker_scan(struct shrinker *shrink, struct shrink_con runtime_pm = xe_shrinker_runtime_pm_get(shrinker, true, 0, can_backup); shrink_flags.purge = false; + lret = xe_shrinker_walk(shrinker->xe, &ctx, shrink_flags, nr_to_scan, &nr_scanned); if (lret >= 0) -- 2.50.1

4 months, 3 weeks

3
2
0 0

[PATCH v2] RDMA/siw: Fix the sendmsg byte count in siw_tcp_sendpages

by Pedro Falcato

Ever since commit c2ff29e99a76 ("siw: Inline do_tcp_sendpages()"), we have been doing this: static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset, size_t size) [...] /* Calculate the number of bytes we need to push, for this page * specifically */ size_t bytes = min_t(size_t, PAGE_SIZE - offset, size); /* If we can't splice it, then copy it in, as normal */ if (!sendpage_ok(page[i])) msg.msg_flags &= ~MSG_SPLICE_PAGES; /* Set the bvec pointing to the page, with len $bytes */ bvec_set_page(&bvec, page[i], bytes, offset); /* Set the iter to $size, aka the size of the whole sendpages (!!!) */ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size); try_page_again: lock_sock(sk); /* Sendmsg with $size size (!!!) */ rv = tcp_sendmsg_locked(sk, &msg, size); This means we've been sending oversized iov_iters and tcp_sendmsg calls for a while. This has a been a benign bug because sendpage_ok() always returned true. With the recent slab allocator changes being slowly introduced into next (that disallow sendpage on large kmalloc allocations), we have recently hit out-of-bounds crashes, due to slight differences in iov_iter behavior between the MSG_SPLICE_PAGES and "regular" copy paths: (MSG_SPLICE_PAGES) skb_splice_from_iter iov_iter_extract_pages iov_iter_extract_bvec_pages uses i->nr_segs to correctly stop in its tracks before OoB'ing everywhere skb_splice_from_iter gets a "short" read (!MSG_SPLICE_PAGES) skb_copy_to_page_nocache copy=iov_iter_count [...] copy_from_iter /* this doesn't help */ if (unlikely(iter->count < len)) len = iter->count; iterate_bvec ... and we run off the bvecs Fix this by properly setting the iov_iter's byte count, plus sending the correct byte count to tcp_sendmsg_locked. Cc: stable(a)vger.kernel.org Fixes: c2ff29e99a76 ("siw: Inline do_tcp_sendpages()") Reported-by: kernel test robot <oliver.sang(a)intel.com> Closes: https://lore.kernel.org/oe-lkp/202507220801.50a7210-lkp@intel.com Reviewed-by: David Howells <dhowells(a)redhat.com> Signed-off-by: Pedro Falcato <pfalcato(a)suse.de> --- v2: - Add David Howells's Rb on the original patch - Remove the offset increment, since it's dead code drivers/infiniband/sw/siw/siw_qp_tx.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/sw/siw/siw_qp_tx.c b/drivers/infiniband/sw/siw/siw_qp_tx.c index 3a08f57d2211..f7dd32c6e5ba 100644 --- a/drivers/infiniband/sw/siw/siw_qp_tx.c +++ b/drivers/infiniband/sw/siw/siw_qp_tx.c @@ -340,18 +340,17 @@ static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset, if (!sendpage_ok(page[i])) msg.msg_flags &= ~MSG_SPLICE_PAGES; bvec_set_page(&bvec, page[i], bytes, offset); - iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size); + iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, bytes); try_page_again: lock_sock(sk); - rv = tcp_sendmsg_locked(sk, &msg, size); + rv = tcp_sendmsg_locked(sk, &msg, bytes); release_sock(sk); if (rv > 0) { size -= rv; sent += rv; if (rv != bytes) { - offset += rv; bytes -= rv; goto try_page_again; } -- 2.50.1

4 months, 3 weeks

4
5
0 0

[PATCH v3] vhost/net: Protect ubufs with rcu read lock in vhost_net_ubuf_put()

by Nikolay Kuratov

When operating on struct vhost_net_ubuf_ref, the following execution sequence is theoretically possible: CPU0 is finalizing DMA operation CPU1 is doing VHOST_NET_SET_BACKEND // ubufs->refcount == 2 vhost_net_ubuf_put() vhost_net_ubuf_put_wait_and_free(oldubufs) vhost_net_ubuf_put_and_wait() vhost_net_ubuf_put() int r = atomic_sub_return(1, &ubufs->refcount); // r = 1 int r = atomic_sub_return(1, &ubufs->refcount); // r = 0 wait_event(ubufs->wait, !atomic_read(&ubufs->refcount)); // no wait occurs here because condition is already true kfree(ubufs); if (unlikely(!r)) wake_up(&ubufs->wait); // use-after-free This leads to use-after-free on ubufs access. This happens because CPU1 skips waiting for wake_up() when refcount is already zero. To prevent that use a read-side RCU critical section in vhost_net_ubuf_put(), as suggested by Hillf Danton. For this lock to take effect, free ubufs with kfree_rcu(). Cc: stable(a)vger.kernel.org Fixes: 0ad8b480d6ee9 ("vhost: fix ref cnt checking deadlock") Reported-by: Andrey Ryabinin <arbn(a)yandex-team.com> Suggested-by: Hillf Danton <hdanton(a)sina.com> Signed-off-by: Nikolay Kuratov <kniv(a)yandex-team.ru> --- v2: * move reinit_completion() into vhost_net_flush(), thanks to Hillf Danton * add Tested-by: Lei Yang * check that usages of put_and_wait() are consistent across LTS kernels v3: * use rcu_read_lock() with kfree_rcu() instead of completion, as suggested by Hillf Danton drivers/vhost/net.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index 6edac0c1ba9b..c6508fe0d5c8 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -99,6 +99,7 @@ struct vhost_net_ubuf_ref { atomic_t refcount; wait_queue_head_t wait; struct vhost_virtqueue *vq; + struct rcu_head rcu; }; #define VHOST_NET_BATCH 64 @@ -250,9 +251,13 @@ vhost_net_ubuf_alloc(struct vhost_virtqueue *vq, bool zcopy) static int vhost_net_ubuf_put(struct vhost_net_ubuf_ref *ubufs) { - int r = atomic_sub_return(1, &ubufs->refcount); + int r; + + rcu_read_lock(); + r = atomic_sub_return(1, &ubufs->refcount); if (unlikely(!r)) wake_up(&ubufs->wait); + rcu_read_unlock(); return r; } @@ -265,7 +270,7 @@ static void vhost_net_ubuf_put_and_wait(struct vhost_net_ubuf_ref *ubufs) static void vhost_net_ubuf_put_wait_and_free(struct vhost_net_ubuf_ref *ubufs) { vhost_net_ubuf_put_and_wait(ubufs); - kfree(ubufs); + kfree_rcu(ubufs, rcu); } static void vhost_net_clear_ubuf_info(struct vhost_net *n) -- 2.34.1

4 months, 3 weeks

1
0
0 0

[PATCH v3 1/3] sched_ext: Mark scx_bpf_cpu_rq as NULL returnable

by Christian Loehle

scx_bpf_cpu_rq() obviously returns NULL on invalid cpu. Mark it as such. While kf_cpu_valid() will trigger scx_ops_error() that leads to the BPF scheduler exiting, this isn't guaranteed to be immediate, allowing for a dereference of a NULL scx_bpf_cpu_rq() return value. Cc: stable(a)vger.kernel.org Fixes: 6203ef73fa5c ("sched/ext: Add BPF function to fetch rq") Signed-off-by: Christian Loehle <christian.loehle(a)arm.com> Acked-by: Andrea Righi <arighi(a)nvidia.com> --- kernel/sched/ext.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c index 7dedc9a16281..3ea3f0f18030 100644 --- a/kernel/sched/ext.c +++ b/kernel/sched/ext.c @@ -7589,7 +7589,7 @@ BTF_ID_FLAGS(func, scx_bpf_get_online_cpumask, KF_ACQUIRE) BTF_ID_FLAGS(func, scx_bpf_put_cpumask, KF_RELEASE) BTF_ID_FLAGS(func, scx_bpf_task_running, KF_RCU) BTF_ID_FLAGS(func, scx_bpf_task_cpu, KF_RCU) -BTF_ID_FLAGS(func, scx_bpf_cpu_rq) +BTF_ID_FLAGS(func, scx_bpf_cpu_rq, KF_RET_NULL) #ifdef CONFIG_CGROUP_SCHED BTF_ID_FLAGS(func, scx_bpf_task_cgroup, KF_RCU | KF_ACQUIRE) #endif -- 2.34.1

4 months, 3 weeks

1
0
0 0

[PATCH] drm/amdgpu: Raven: don't allow mixing GTT and VRAM

by Brian Geffon

Commit 81d0bcf99009 ("drm/amdgpu: make display pinning more flexible (v2)") allowed for newer ASICs to mix GTT and VRAM, this change also noted that some older boards, such as Stoney and Carrizo do not support this. It appears that at least one additional ASIC does not support this which is Raven. We observed this issue when migrating a device from a 5.4 to 6.6 kernel and have confirmed that Raven also needs to be excluded from mixing GTT and VRAM. Fixes: 81d0bcf99009 ("drm/amdgpu: make display pinning more flexible (v2)") Cc: Luben Tuikov <luben.tuikov(a)amd.com> Cc: Christian König <christian.koenig(a)amd.com> Cc: Alex Deucher <alexander.deucher(a)amd.com> Cc: stable(a)vger.kernel.org # 6.1+ Tested-by: Thadeu Lima de Souza Cascardo <cascardo(a)igalia.com> Signed-off-by: Brian Geffon <bgeffon(a)google.com> --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 73403744331a..5d7f13e25b7c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -1545,7 +1545,8 @@ uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev, uint32_t domain) { if ((domain == (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) && - ((adev->asic_type == CHIP_CARRIZO) || (adev->asic_type == CHIP_STONEY))) { + ((adev->asic_type == CHIP_CARRIZO) || (adev->asic_type == CHIP_STONEY) || + (adev->asic_type == CHIP_RAVEN))) { domain = AMDGPU_GEM_DOMAIN_VRAM; if (adev->gmc.real_vram_size <= AMDGPU_SG_THRESHOLD) domain = AMDGPU_GEM_DOMAIN_GTT; -- 2.50.0.727.gbf7dc18ff4-goog

4 months, 3 weeks

5
16
0 0

[PATCH v2] vhost/net: Replace wait_queue with completion in ubufs reference

by Nikolay Kuratov

When operating on struct vhost_net_ubuf_ref, the following execution sequence is theoretically possible: CPU0 is finalizing DMA operation CPU1 is doing VHOST_NET_SET_BACKEND // &ubufs->refcount == 2 vhost_net_ubuf_put() vhost_net_ubuf_put_wait_and_free(oldubufs) vhost_net_ubuf_put_and_wait() vhost_net_ubuf_put() int r = atomic_sub_return(1, &ubufs->refcount); // r = 1 int r = atomic_sub_return(1, &ubufs->refcount); // r = 0 wait_event(ubufs->wait, !atomic_read(&ubufs->refcount)); // no wait occurs here because condition is already true kfree(ubufs); if (unlikely(!r)) wake_up(&ubufs->wait); // use-after-free This leads to use-after-free on ubufs access. This happens because CPU1 skips waiting for wake_up() when refcount is already zero. To prevent that use a completion instead of wait_queue as the ubufs notification mechanism. wait_for_completion() guarantees that there will be complete() call prior to its return. We also need to reinit completion in vhost_net_flush(), because refcnt == 0 does not mean freeing in that case. Cc: stable(a)vger.kernel.org Fixes: 0ad8b480d6ee9 ("vhost: fix ref cnt checking deadlock") Reported-by: Andrey Ryabinin <arbn(a)yandex-team.com> Suggested-by: Andrey Smetanin <asmetanin(a)yandex-team.ru> Suggested-by: Hillf Danton <hdanton(a)sina.com> Tested-by: Lei Yang <leiyang(a)redhat.com> (v1) Signed-off-by: Nikolay Kuratov <kniv(a)yandex-team.ru> --- v2: * move reinit_completion() into vhost_net_flush(), thanks to Hillf Danton * add Tested-by: Lei Yang * check that usages of put_and_wait() are consistent across LTS kernels drivers/vhost/net.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index 7cbfc7d718b3..69e1bfb9627e 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -94,7 +94,7 @@ struct vhost_net_ubuf_ref { * >1: outstanding ubufs */ atomic_t refcount; - wait_queue_head_t wait; + struct completion wait; struct vhost_virtqueue *vq; }; @@ -240,7 +240,7 @@ vhost_net_ubuf_alloc(struct vhost_virtqueue *vq, bool zcopy) if (!ubufs) return ERR_PTR(-ENOMEM); atomic_set(&ubufs->refcount, 1); - init_waitqueue_head(&ubufs->wait); + init_completion(&ubufs->wait); ubufs->vq = vq; return ubufs; } @@ -249,14 +249,14 @@ static int vhost_net_ubuf_put(struct vhost_net_ubuf_ref *ubufs) { int r = atomic_sub_return(1, &ubufs->refcount); if (unlikely(!r)) - wake_up(&ubufs->wait); + complete_all(&ubufs->wait); return r; } static void vhost_net_ubuf_put_and_wait(struct vhost_net_ubuf_ref *ubufs) { vhost_net_ubuf_put(ubufs); - wait_event(ubufs->wait, !atomic_read(&ubufs->refcount)); + wait_for_completion(&ubufs->wait); } static void vhost_net_ubuf_put_wait_and_free(struct vhost_net_ubuf_ref *ubufs) @@ -1381,6 +1381,7 @@ static void vhost_net_flush(struct vhost_net *n) mutex_lock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex); n->tx_flush = false; atomic_set(&n->vqs[VHOST_NET_VQ_TX].ubufs->refcount, 1); + reinit_completion(&n->vqs[VHOST_NET_VQ_TX].ubufs->wait); mutex_unlock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex); } } -- 2.34.1

4 months, 3 weeks

2
1
0 0

[PATCH] rust: faux: fix C header link

by Miguel Ojeda

Starting with Rust 1.91.0 (expected 2025-10-30), `rustdoc` has improved some false negatives around intra-doc links [1], and it found a broken intra-doc link we currently have: error: unresolved link to `include/linux/device/faux.h` --> rust/kernel/faux.rs:7:17 | 7 | //! C header: [`include/linux/device/faux.h`] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ no item named `include/linux/device/faux.h` in scope | = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` = note: `-D rustdoc::broken-intra-doc-links` implied by `-D warnings` = help: to override `-D warnings` add `#[allow(rustdoc::broken_intra_doc_links)]` Our `srctree/` C header links are not intra-doc links, thus they need the link destination. Thus fix it. Cc: stable(a)vger.kernel.org Link: https://github.com/rust-lang/rust/pull/132748 [1] Fixes: 78418f300d39 ("rust/kernel: Add faux device bindings") Signed-off-by: Miguel Ojeda <ojeda(a)kernel.org> --- It may have been in 1.90, but the beta branch does not have it, and the rollup PR says 1.91, unlike the PR itself, so I picked 1.91. It happened just after the version bump to 1.91, so it may have to do with that. rust/kernel/faux.rs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/rust/kernel/faux.rs b/rust/kernel/faux.rs index 7a906099993f..7fe2dd197e37 100644 --- a/rust/kernel/faux.rs +++ b/rust/kernel/faux.rs @@ -4,7 +4,7 @@ //! //! This module provides bindings for working with faux devices in kernel modules. //! -//! C header: [`include/linux/device/faux.h`] +//! C header: [`include/linux/device/faux.h`](srctree/include/linux/device/faux.h) use crate::{bindings, device, error::code::*, prelude::*}; use core::ptr::{addr_of_mut, null, null_mut, NonNull}; base-commit: d2eedaa3909be9102d648a4a0a50ccf64f96c54f -- 2.50.1

4 months, 3 weeks

3
2
0 0

[PATCH v6] riscv: hwprobe: Fix stale vDSO data for late-initialized keys at boot

by Jingwei Wang

The hwprobe vDSO data for some keys, like MISALIGNED_VECTOR_PERF, is determined by an asynchronous kthread. This can create a race condition where the kthread finishes after the vDSO data has already been populated, causing userspace to read stale values. To fix this race, a new 'ready' flag is added to the vDSO data, initialized to 'false' during late_initcall. This flag is checked by both the vDSO's user-space code and the riscv_hwprobe syscall. The syscall serves as a one-time gate, using a completion to wait for any pending probes before populating the data and setting the flag to 'true', thus ensuring userspace reads fresh values on its first request. Reported-by: Tsukasa OI <research_trasio(a)irq.a4lg.com> Closes: https://lore.kernel.org/linux-riscv/760d637b-b13b-4518-b6bf-883d55d44e7f@ir… Fixes: e7c9d66e313b ("RISC-V: Report vector unaligned access speed hwprobe") Cc: Palmer Dabbelt <palmer(a)dabbelt.com> Cc: Alexandre Ghiti <alexghiti(a)rivosinc.com> Cc: Olof Johansson <olof(a)lixom.net> Cc: stable(a)vger.kernel.org Co-developed-by: Palmer Dabbelt <palmer(a)dabbelt.com> Signed-off-by: Jingwei Wang <wangjingwei(a)iscas.ac.cn> --- Changes in v6: - Based on Palmer's feedback, reworked the synchronization to be on-demand, deferring the wait until the first hwprobe syscall via a 'ready' flag. This avoids the boot-time regression from v5's approach. Changes in v5: - Reworked the synchronization logic to a robust "sentinel-count" pattern based on feedback from Alexandre. - Fixed a "multiple definition" linker error for nommu builds by changing the header-file stub functions to `static inline`, as pointed out by Olof. - Updated the commit message to better explain the rationale for moving the vDSO initialization to `late_initcall`. Changes in v4: - Reworked the synchronization mechanism based on feedback from Palmer and Alexandre. - Instead of a post-hoc refresh, this version introduces a robust completion-based framework using an atomic counter to ensure async probes are finished before populating the vDSO. - Moved the vdso data initialization to a late_initcall to avoid impacting boot time. Changes in v3: - Retained existing blank line. Changes in v2: - Addressed feedback from Yixun's regarding #ifdef CONFIG_MMU usage. - Updated commit message to provide a high-level summary. - Added Fixes tag for commit e7c9d66e313b. v1: https://lore.kernel.org/linux-riscv/20250521052754.185231-1-wangjingwei@isc… arch/riscv/include/asm/hwprobe.h | 8 ++- arch/riscv/include/asm/vdso/arch_data.h | 6 ++ arch/riscv/kernel/sys_hwprobe.c | 71 ++++++++++++++++++---- arch/riscv/kernel/unaligned_access_speed.c | 9 ++- arch/riscv/kernel/vdso/hwprobe.c | 2 +- 5 files changed, 79 insertions(+), 17 deletions(-) diff --git a/arch/riscv/include/asm/hwprobe.h b/arch/riscv/include/asm/hwprobe.h index 7fe0a379474ae2c6..3b2888126e659ea1 100644 --- a/arch/riscv/include/asm/hwprobe.h +++ b/arch/riscv/include/asm/hwprobe.h @@ -40,5 +40,11 @@ static inline bool riscv_hwprobe_pair_cmp(struct riscv_hwprobe *pair, return pair->value == other_pair->value; } - +#ifdef CONFIG_MMU +void riscv_hwprobe_register_async_probe(void); +void riscv_hwprobe_complete_async_probe(void); +#else +static inline void riscv_hwprobe_register_async_probe(void) {} +static inline void riscv_hwprobe_complete_async_probe(void) {} +#endif #endif diff --git a/arch/riscv/include/asm/vdso/arch_data.h b/arch/riscv/include/asm/vdso/arch_data.h index da57a3786f7a53c8..88b37af55175129b 100644 --- a/arch/riscv/include/asm/vdso/arch_data.h +++ b/arch/riscv/include/asm/vdso/arch_data.h @@ -12,6 +12,12 @@ struct vdso_arch_data { /* Boolean indicating all CPUs have the same static hwprobe values. */ __u8 homogeneous_cpus; + + /* + * A gate to check and see if the hwprobe data is actually ready, as + * probing is deferred to avoid boot slowdowns. + */ + __u8 ready; }; #endif /* __RISCV_ASM_VDSO_ARCH_DATA_H */ diff --git a/arch/riscv/kernel/sys_hwprobe.c b/arch/riscv/kernel/sys_hwprobe.c index 0b170e18a2beba57..fecb6790fa88e96c 100644 --- a/arch/riscv/kernel/sys_hwprobe.c +++ b/arch/riscv/kernel/sys_hwprobe.c @@ -5,6 +5,8 @@ * more details. */ #include <linux/syscalls.h> +#include <linux/completion.h> +#include <linux/atomic.h> #include <asm/cacheflush.h> #include <asm/cpufeature.h> #include <asm/hwprobe.h> @@ -452,28 +454,36 @@ static int hwprobe_get_cpus(struct riscv_hwprobe __user *pairs, return 0; } -static int do_riscv_hwprobe(struct riscv_hwprobe __user *pairs, - size_t pair_count, size_t cpusetsize, - unsigned long __user *cpus_user, - unsigned int flags) -{ - if (flags & RISCV_HWPROBE_WHICH_CPUS) - return hwprobe_get_cpus(pairs, pair_count, cpusetsize, - cpus_user, flags); +#ifdef CONFIG_MMU - return hwprobe_get_values(pairs, pair_count, cpusetsize, - cpus_user, flags); +static DECLARE_COMPLETION(boot_probes_done); +static atomic_t pending_boot_probes = ATOMIC_INIT(1); + +void riscv_hwprobe_register_async_probe(void) +{ + atomic_inc(&pending_boot_probes); } -#ifdef CONFIG_MMU +void riscv_hwprobe_complete_async_probe(void) +{ + if (atomic_dec_and_test(&pending_boot_probes)) + complete(&boot_probes_done); +} -static int __init init_hwprobe_vdso_data(void) +static int complete_hwprobe_vdso_data(void) { struct vdso_arch_data *avd = vdso_k_arch_data; u64 id_bitsmash = 0; struct riscv_hwprobe pair; int key; + /* We've probably already produced these values. */ + if (likely(avd->ready)) + return 0; + + if (unlikely(!atomic_dec_and_test(&pending_boot_probes))) + wait_for_completion(&boot_probes_done); + /* * Initialize vDSO data with the answers for the "all CPUs" case, to * save a syscall in the common case. @@ -501,13 +511,48 @@ static int __init init_hwprobe_vdso_data(void) * vDSO should defer to the kernel for exotic cpu masks. */ avd->homogeneous_cpus = id_bitsmash != 0 && id_bitsmash != -1; + + /* + * Make sure all the VDSO values are visible before we look at them. + * This pairs with the implicit "no speculativly visible accesses" + * barrier in the VDSO hwprobe code. + */ + smp_wmb(); + avd->ready = true; + return 0; +} + +static int __init init_hwprobe_vdso_data(void) +{ + struct vdso_arch_data *avd = vdso_k_arch_data; + + /* + * Prevent the vDSO cached values from being used, as they're not ready + * yet. + */ + avd->ready = false; return 0; } -arch_initcall_sync(init_hwprobe_vdso_data); +late_initcall(init_hwprobe_vdso_data); #endif /* CONFIG_MMU */ +static int do_riscv_hwprobe(struct riscv_hwprobe __user *pairs, + size_t pair_count, size_t cpusetsize, + unsigned long __user *cpus_user, + unsigned int flags) +{ + complete_hwprobe_vdso_data(); + + if (flags & RISCV_HWPROBE_WHICH_CPUS) + return hwprobe_get_cpus(pairs, pair_count, cpusetsize, + cpus_user, flags); + + return hwprobe_get_values(pairs, pair_count, cpusetsize, + cpus_user, flags); +} + SYSCALL_DEFINE5(riscv_hwprobe, struct riscv_hwprobe __user *, pairs, size_t, pair_count, size_t, cpusetsize, unsigned long __user *, cpus, unsigned int, flags) diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c index ae2068425fbcd207..4b8ad2673b0f7470 100644 --- a/arch/riscv/kernel/unaligned_access_speed.c +++ b/arch/riscv/kernel/unaligned_access_speed.c @@ -379,6 +379,7 @@ static void check_vector_unaligned_access(struct work_struct *work __always_unus static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused) { schedule_on_each_cpu(check_vector_unaligned_access); + riscv_hwprobe_complete_async_probe(); return 0; } @@ -473,8 +474,12 @@ static int __init check_unaligned_access_all_cpus(void) per_cpu(vector_misaligned_access, cpu) = unaligned_vector_speed_param; } else if (!check_vector_unaligned_access_emulated_all_cpus() && IS_ENABLED(CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS)) { - kthread_run(vec_check_unaligned_access_speed_all_cpus, - NULL, "vec_check_unaligned_access_speed_all_cpus"); + riscv_hwprobe_register_async_probe(); + if (IS_ERR(kthread_run(vec_check_unaligned_access_speed_all_cpus, + NULL, "vec_check_unaligned_access_speed_all_cpus"))) { + pr_warn("Failed to create vec_unalign_check kthread\n"); + riscv_hwprobe_complete_async_probe(); + } } /* diff --git a/arch/riscv/kernel/vdso/hwprobe.c b/arch/riscv/kernel/vdso/hwprobe.c index 2ddeba6c68dda09b..bf77b4c1d2d8e803 100644 --- a/arch/riscv/kernel/vdso/hwprobe.c +++ b/arch/riscv/kernel/vdso/hwprobe.c @@ -27,7 +27,7 @@ static int riscv_vdso_get_values(struct riscv_hwprobe *pairs, size_t pair_count, * homogeneous, then this function can handle requests for arbitrary * masks. */ - if ((flags != 0) || (!all_cpus && !avd->homogeneous_cpus)) + if ((flags != 0) || (!all_cpus && !avd->homogeneous_cpus) || unlikely(!avd->ready)) return riscv_hwprobe(pairs, pair_count, cpusetsize, cpus, flags); /* This is something we can handle, fill out the pairs. */ -- 2.50.1

4 months, 3 weeks

2
2
0 0

[PATCH v2] fs/adfs: bigdir: Restore EIO errno return when checkbyte mismatch

by Zhen Ni

The error path in adfs_fplus_read() prints an error message via adfs_error() but incorrectly returns success (0) to the caller. This occurs because the 'ret' variable remains set to 0 from the earlier successful call to adfs_fplus_validate_tail(). Fix by setting 'ret = -EIO' before jumping to the error exit. This issue was detected by smatch static analysis: warning: fs/adfs/dir_fplus.c:146: adfs_fplus_read() warn: missing error code 'ret'. Fixes: d79288b4f61b ("fs/adfs: bigdir: calculate and validate directory checkbyte") Cc: stable(a)vger.kernel.org Signed-off-by: Zhen Ni <zhen.ni(a)easystack.cn> --- Changes in v2: - Add tags of 'Fixes' and 'Cc' in commit message - Added error description and the corresponding fix in commit message --- fs/adfs/dir_fplus.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/adfs/dir_fplus.c b/fs/adfs/dir_fplus.c index 4a15924014da..4334279409b2 100644 --- a/fs/adfs/dir_fplus.c +++ b/fs/adfs/dir_fplus.c @@ -143,6 +143,7 @@ static int adfs_fplus_read(struct super_block *sb, u32 indaddr, if (adfs_fplus_checkbyte(dir) != t->bigdircheckbyte) { adfs_error(sb, "dir %06x checkbyte mismatch\n", indaddr); + ret = -EIO; goto out; } -- 2.20.1

4 months, 3 weeks

1
0
0 0

[PATCH] drm/xe: Defer buffer object shrinker write-backs and GPU waits

by Thomas Hellström

When the xe buffer-object shrinker allows GPU waits and write-back, (typically from kswapd), perform multilpe passes, skipping subsequent passes if the shrinker number of scanned objects target is reached. 1) Without GPU waits and write-back 2) Without write-back 3) With both GPU-waits and write-back This is to avoid stalls and costly write- and readbacks unless they are really necessary. Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/5557#note_3035136 Fixes: 00c8efc3180f ("drm/xe: Add a shrinker for xe bos") Cc: <stable(a)vger.kernel.org> # v6.15+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_shrinker.c | 51 +++++++++++++++++++++++++++++--- 1 file changed, 47 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_shrinker.c b/drivers/gpu/drm/xe/xe_shrinker.c index 1c3c04d52f55..bc3439bd4450 100644 --- a/drivers/gpu/drm/xe/xe_shrinker.c +++ b/drivers/gpu/drm/xe/xe_shrinker.c @@ -54,10 +54,10 @@ xe_shrinker_mod_pages(struct xe_shrinker *shrinker, long shrinkable, long purgea write_unlock(&shrinker->lock); } -static s64 xe_shrinker_walk(struct xe_device *xe, - struct ttm_operation_ctx *ctx, - const struct xe_bo_shrink_flags flags, - unsigned long to_scan, unsigned long *scanned) +static s64 __xe_shrinker_walk(struct xe_device *xe, + struct ttm_operation_ctx *ctx, + const struct xe_bo_shrink_flags flags, + unsigned long to_scan, unsigned long *scanned) { unsigned int mem_type; s64 freed = 0, lret; @@ -93,6 +93,48 @@ static s64 xe_shrinker_walk(struct xe_device *xe, return freed; } +/* + * Try shrinking idle objects without writeback first, then if not sufficient, + * try also non-idle objects and finally if that's not sufficient either, + * add writeback. This avoids stalls and explicit writebacks with light or + * moderate memory pressure. + */ +static s64 xe_shrinker_walk(struct xe_device *xe, + struct ttm_operation_ctx *ctx, + const struct xe_bo_shrink_flags flags, + unsigned long to_scan, unsigned long *scanned) +{ + bool no_wait_gpu = true; + struct xe_bo_shrink_flags save_flags = flags; + s64 lret, freed; + + swap(no_wait_gpu, ctx->no_wait_gpu); + save_flags.writeback = false; + lret = __xe_shrinker_walk(xe, ctx, save_flags, to_scan, scanned); + swap(no_wait_gpu, ctx->no_wait_gpu); + if (lret < 0 || *scanned >= to_scan) + return lret; + + freed = lret; + if (!ctx->no_wait_gpu) { + lret = __xe_shrinker_walk(xe, ctx, save_flags, to_scan, scanned); + if (lret < 0) + return lret; + freed += lret; + } + if (*scanned >= to_scan) + return freed; + + if (flags.writeback) { + lret = __xe_shrinker_walk(xe, ctx, flags, to_scan, scanned); + if (lret < 0) + return lret; + freed += lret; + } + + return freed; +} + static unsigned long xe_shrinker_count(struct shrinker *shrink, struct shrink_control *sc) { @@ -199,6 +241,7 @@ static unsigned long xe_shrinker_scan(struct shrinker *shrink, struct shrink_con runtime_pm = xe_shrinker_runtime_pm_get(shrinker, true, 0, can_backup); shrink_flags.purge = false; + lret = xe_shrinker_walk(shrinker->xe, &ctx, shrink_flags, nr_to_scan, &nr_scanned); if (lret >= 0) -- 2.50.1

4 months, 3 weeks

2
2
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror