May 2024 - Linux-stable-mirror

[PATCH v3] mm: Fix race between __split_huge_pmd_locked() and GUP-fast

by Ryan Roberts

__split_huge_pmd_locked() can be called for a present THP, devmap or (non-present) migration entry. It calls pmdp_invalidate() unconditionally on the pmdp and only determines if it is present or not based on the returned old pmd. This is a problem for the migration entry case because pmd_mkinvalid(), called by pmdp_invalidate() must only be called for a present pmd. On arm64 at least, pmd_mkinvalid() will mark the pmd such that any future call to pmd_present() will return true. And therefore any lockless pgtable walker could see the migration entry pmd in this state and start interpretting the fields as if it were present, leading to BadThings (TM). GUP-fast appears to be one such lockless pgtable walker. x86 does not suffer the above problem, but instead pmd_mkinvalid() will corrupt the offset field of the swap entry within the swap pte. See link below for discussion of that problem. Fix all of this by only calling pmdp_invalidate() for a present pmd. And for good measure let's add a warning to all implementations of pmdp_invalidate[_ad](). I've manually reviewed all other pmdp_invalidate[_ad]() call sites and believe all others to be conformant. This is a theoretical bug found during code review. I don't have any test case to trigger it in practice. Cc: stable(a)vger.kernel.org Link: https://lore.kernel.org/all/0dd7827a-6334-439a-8fd0-43c98e6af22b@arm.com/ Fixes: 84c3fc4e9c56 ("mm: thp: check pmd migration entry in common path") Signed-off-by: Ryan Roberts <ryan.roberts(a)arm.com> --- Right v3; this goes back to the original approach in v1 to fix core-mm rather than push the fix into arm64, since we discovered that x86 can't handle pmd_mkinvalid() being called for non-present pmds either. I'm pulling in more arch maintainers because this version adds some warnings in arch code to help spot incorrect usage. Although Catalin had already accepted v2 (fixing arm64) [2] into for-next/fixes, he's agreed to either remove or revert it. Changes since v1 [1] ==================== - Improve pmdp_mkinvalid() docs to make it clear it can only be called for present pmd (per JohnH, Zi Yan) - Added warnings to arch overrides of pmdp_invalidate[_ad]() (per Zi Yan) - Moved comment next to new location of pmpd_invalidate() (per Zi Yan) [1] https://lore.kernel.org/linux-mm/20240425170704.3379492-1-ryan.roberts@arm.… [2] https://lore.kernel.org/all/20240430133138.732088-1-ryan.roberts@arm.com/ Thanks, Ryan Documentation/mm/arch_pgtable_helpers.rst | 6 ++- arch/powerpc/mm/book3s64/pgtable.c | 1 + arch/s390/include/asm/pgtable.h | 4 +- arch/sparc/mm/tlb.c | 1 + arch/x86/mm/pgtable.c | 2 + mm/huge_memory.c | 49 ++++++++++++----------- mm/pgtable-generic.c | 2 + 7 files changed, 39 insertions(+), 26 deletions(-) diff --git a/Documentation/mm/arch_pgtable_helpers.rst b/Documentation/mm/arch_pgtable_helpers.rst index 2466d3363af7..ad50ca6f495e 100644 --- a/Documentation/mm/arch_pgtable_helpers.rst +++ b/Documentation/mm/arch_pgtable_helpers.rst @@ -140,7 +140,8 @@ PMD Page Table Helpers +---------------------------+--------------------------------------------------+ | pmd_swp_clear_soft_dirty | Clears a soft dirty swapped PMD | +---------------------------+--------------------------------------------------+ -| pmd_mkinvalid | Invalidates a mapped PMD [1] | +| pmd_mkinvalid | Invalidates a present PMD; do not call for | +| | non-present PMD [1] | +---------------------------+--------------------------------------------------+ | pmd_set_huge | Creates a PMD huge mapping | +---------------------------+--------------------------------------------------+ @@ -196,7 +197,8 @@ PUD Page Table Helpers +---------------------------+--------------------------------------------------+ | pud_mkdevmap | Creates a ZONE_DEVICE mapped PUD | +---------------------------+--------------------------------------------------+ -| pud_mkinvalid | Invalidates a mapped PUD [1] | +| pud_mkinvalid | Invalidates a present PUD; do not call for | +| | non-present PUD [1] | +---------------------------+--------------------------------------------------+ | pud_set_huge | Creates a PUD huge mapping | +---------------------------+--------------------------------------------------+ diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c index 83823db3488b..2975ea0841ba 100644 --- a/arch/powerpc/mm/book3s64/pgtable.c +++ b/arch/powerpc/mm/book3s64/pgtable.c @@ -170,6 +170,7 @@ pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, { unsigned long old_pmd; + VM_WARN_ON_ONCE(!pmd_present(*pmdp)); old_pmd = pmd_hugepage_update(vma->vm_mm, address, pmdp, _PAGE_PRESENT, _PAGE_INVALID); flush_pmd_tlb_range(vma, address, address + HPAGE_PMD_SIZE); return __pmd(old_pmd); diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h index 60950e7a25f5..480bea44559d 100644 --- a/arch/s390/include/asm/pgtable.h +++ b/arch/s390/include/asm/pgtable.h @@ -1768,8 +1768,10 @@ static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma, static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long addr, pmd_t *pmdp) { - pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID); + pmd_t pmd; + VM_WARN_ON_ONCE(!pmd_present(*pmdp)); + pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID); return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd); } diff --git a/arch/sparc/mm/tlb.c b/arch/sparc/mm/tlb.c index b44d79d778c7..ef69127d7e5e 100644 --- a/arch/sparc/mm/tlb.c +++ b/arch/sparc/mm/tlb.c @@ -249,6 +249,7 @@ pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, { pmd_t old, entry; + VM_WARN_ON_ONCE(!pmd_present(*pmdp)); entry = __pmd(pmd_val(*pmdp) & ~_PAGE_VALID); old = pmdp_establish(vma, address, pmdp, entry); flush_tlb_range(vma, address, address + HPAGE_PMD_SIZE); diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c index d007591b8059..103cbccf1d7d 100644 --- a/arch/x86/mm/pgtable.c +++ b/arch/x86/mm/pgtable.c @@ -631,6 +631,8 @@ int pmdp_clear_flush_young(struct vm_area_struct *vma, pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address, pmd_t *pmdp) { + VM_WARN_ON_ONCE(!pmd_present(*pmdp)); + /* * No flush is necessary. Once an invalid PTE is established, the PTE's * access and dirty bits cannot be updated. diff --git a/mm/huge_memory.c b/mm/huge_memory.c index 89f58c7603b2..dd1fc105f70b 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -2493,32 +2493,11 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd, return __split_huge_zero_page_pmd(vma, haddr, pmd); } - /* - * Up to this point the pmd is present and huge and userland has the - * whole access to the hugepage during the split (which happens in - * place). If we overwrite the pmd with the not-huge version pointing - * to the pte here (which of course we could if all CPUs were bug - * free), userland could trigger a small page size TLB miss on the - * small sized TLB while the hugepage TLB entry is still established in - * the huge TLB. Some CPU doesn't like that. - * See http://support.amd.com/TechDocs/41322_10h_Rev_Gd.pdf, Erratum - * 383 on page 105. Intel should be safe but is also warns that it's - * only safe if the permission and cache attributes of the two entries - * loaded in the two TLB is identical (which should be the case here). - * But it is generally safer to never allow small and huge TLB entries - * for the same virtual address to be loaded simultaneously. So instead - * of doing "pmd_populate(); flush_pmd_tlb_range();" we first mark the - * current pmd notpresent (atomically because here the pmd_trans_huge - * must remain set at all times on the pmd until the split is complete - * for this pmd), then we flush the SMP TLB and finally we write the - * non-huge version of the pmd entry with pmd_populate. - */ - old_pmd = pmdp_invalidate(vma, haddr, pmd); - - pmd_migration = is_pmd_migration_entry(old_pmd); + pmd_migration = is_pmd_migration_entry(*pmd); if (unlikely(pmd_migration)) { swp_entry_t entry; + old_pmd = *pmd; entry = pmd_to_swp_entry(old_pmd); page = pfn_swap_entry_to_page(entry); write = is_writable_migration_entry(entry); @@ -2529,6 +2508,30 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd, soft_dirty = pmd_swp_soft_dirty(old_pmd); uffd_wp = pmd_swp_uffd_wp(old_pmd); } else { + /* + * Up to this point the pmd is present and huge and userland has + * the whole access to the hugepage during the split (which + * happens in place). If we overwrite the pmd with the not-huge + * version pointing to the pte here (which of course we could if + * all CPUs were bug free), userland could trigger a small page + * size TLB miss on the small sized TLB while the hugepage TLB + * entry is still established in the huge TLB. Some CPU doesn't + * like that. See + * http://support.amd.com/TechDocs/41322_10h_Rev_Gd.pdf, Erratum + * 383 on page 105. Intel should be safe but is also warns that + * it's only safe if the permission and cache attributes of the + * two entries loaded in the two TLB is identical (which should + * be the case here). But it is generally safer to never allow + * small and huge TLB entries for the same virtual address to be + * loaded simultaneously. So instead of doing "pmd_populate(); + * flush_pmd_tlb_range();" we first mark the current pmd + * notpresent (atomically because here the pmd_trans_huge must + * remain set at all times on the pmd until the split is + * complete for this pmd), then we flush the SMP TLB and finally + * we write the non-huge version of the pmd entry with + * pmd_populate. + */ + old_pmd = pmdp_invalidate(vma, haddr, pmd); page = pmd_page(old_pmd); folio = page_folio(page); if (pmd_dirty(old_pmd)) { diff --git a/mm/pgtable-generic.c b/mm/pgtable-generic.c index 4fcd959dcc4d..a78a4adf711a 100644 --- a/mm/pgtable-generic.c +++ b/mm/pgtable-generic.c @@ -198,6 +198,7 @@ pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp) pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, pmd_t *pmdp) { + VM_WARN_ON_ONCE(!pmd_present(*pmdp)); pmd_t old = pmdp_establish(vma, address, pmdp, pmd_mkinvalid(*pmdp)); flush_pmd_tlb_range(vma, address, address + HPAGE_PMD_SIZE); return old; @@ -208,6 +209,7 @@ pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address, pmd_t *pmdp) { + VM_WARN_ON_ONCE(!pmd_present(*pmdp)); return pmdp_invalidate(vma, address, pmdp); } #endif -- 2.25.1

10 months, 1 week

5
8
0 0

[PATCH 1/1] nvme-pci: add NVME_QUIRK_BOGUS_NID for Samsung PM173X

by Saeed Mirzamohammadi

This adds a quirk to fix the Samsung PM1733a and PM173X reporting bogus eui64 so they are not marked as "non globally unique" duplicates. Cc: <stable(a)vger.kernel.org> Signed-off-by: Saeed Mirzamohammadi <saeed.mirzamohammadi(a)oracle.com> --- drivers/nvme/host/pci.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c index 5b95c94ee40f2..c0b1caba1c893 100644 --- a/drivers/nvme/host/pci.c +++ b/drivers/nvme/host/pci.c @@ -3359,6 +3359,10 @@ static const struct pci_device_id nvme_id_table[] = { .driver_data = NVME_QUIRK_DELAY_BEFORE_CHK_RDY | NVME_QUIRK_DISABLE_WRITE_ZEROES| NVME_QUIRK_IGNORE_DEV_SUBNQN, }, + { PCI_DEVICE(0x144d, 0xa824), /* Samsung PM173X */ + .driver_data = NVME_QUIRK_BOGUS_NID, }, + { PCI_DEVICE(0x144d, 0xa825), /* Samsung PM1733a */ + .driver_data = NVME_QUIRK_BOGUS_NID, }, { PCI_DEVICE(0x1987, 0x5012), /* Phison E12 */ .driver_data = NVME_QUIRK_BOGUS_NID, }, { PCI_DEVICE(0x1987, 0x5016), /* Phison E16 */ -- 2.39.2

10 months, 1 week

6
13
0 0

[PATCH 6.1 1/1] tls: fix race between tx work scheduling and socket close

by Lee Jones

From: Jakub Kicinski <kuba(a)kernel.org> [ Upstream commit e01e3934a1b2d122919f73bc6ddbe1cdafc4bbdb ] Similarly to previous commit, the submitting thread (recvmsg/sendmsg) may exit as soon as the async crypto handler calls complete(). Reorder scheduling the work before calling complete(). This seems more logical in the first place, as it's the inverse order of what the submitting thread will do. Reported-by: valis <sec(a)valis.email> Fixes: a42055e8d2c3 ("net/tls: Add support for async encryption of records for performance") Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> Reviewed-by: Simon Horman <horms(a)kernel.org> Reviewed-by: Sabrina Dubroca <sd(a)queasysnail.net> Signed-off-by: David S. Miller <davem(a)davemloft.net> Signed-off-by: Sasha Levin <sashal(a)kernel.org> (cherry picked from commit 6db22d6c7a6dc914b12c0469b94eb639b6a8a146) [Lee: Fixed merge-conflict in Stable branches linux-6.1.y and older] Signed-off-by: Lee Jones <lee(a)kernel.org> --- net/tls/tls_sw.c | 16 ++++++---------- 1 file changed, 6 insertions(+), 10 deletions(-) diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c index 2bd27b77769cb..d53587ff9ddea 100644 --- a/net/tls/tls_sw.c +++ b/net/tls/tls_sw.c @@ -449,7 +449,6 @@ static void tls_encrypt_done(crypto_completion_data_t *data, int err) struct scatterlist *sge; struct sk_msg *msg_en; struct tls_rec *rec; - bool ready = false; struct sock *sk; rec = container_of(aead_req, struct tls_rec, aead_req); @@ -486,19 +485,16 @@ static void tls_encrypt_done(crypto_completion_data_t *data, int err) /* If received record is at head of tx_list, schedule tx */ first_rec = list_first_entry(&ctx->tx_list, struct tls_rec, list); - if (rec == first_rec) - ready = true; + if (rec == first_rec) { + /* Schedule the transmission */ + if (!test_and_set_bit(BIT_TX_SCHEDULED, + &ctx->tx_bitmask)) + schedule_delayed_work(&ctx->tx_work.work, 1); + } } if (atomic_dec_and_test(&ctx->encrypt_pending)) complete(&ctx->async_wait.completion); - - if (!ready) - return; - - /* Schedule the transmission */ - if (!test_and_set_bit(BIT_TX_SCHEDULED, &ctx->tx_bitmask)) - schedule_delayed_work(&ctx->tx_work.work, 1); } static int tls_encrypt_async_wait(struct tls_sw_context_tx *ctx) -- 2.44.0.278.ge034bb2e1d-goog

10 months, 3 weeks

5
8
0 0

[PATCH v4 RESEND] lockdep: fix deadlock issue between lockdep and rcu

by Carlos Llamas

From: Zhiguo Niu <zhiguo.niu(a)unisoc.com> There is a deadlock scenario between lockdep and rcu when rcu nocb feature is enabled, just as following call stack: rcuop/x -000|queued_spin_lock_slowpath(lock = 0xFFFFFF817F2A8A80, val = ?) -001|queued_spin_lock(inline) // try to hold nocb_gp_lock -001|do_raw_spin_lock(lock = 0xFFFFFF817F2A8A80) -002|__raw_spin_lock_irqsave(inline) -002|_raw_spin_lock_irqsave(lock = 0xFFFFFF817F2A8A80) -003|wake_nocb_gp_defer(inline) -003|__call_rcu_nocb_wake(rdp = 0xFFFFFF817F30B680) -004|__call_rcu_common(inline) -004|call_rcu(head = 0xFFFFFFC082EECC28, func = ?) -005|call_rcu_zapped(inline) -005|free_zapped_rcu(ch = ?)// hold graph lock -006|rcu_do_batch(rdp = 0xFFFFFF817F245680) -007|nocb_cb_wait(inline) -007|rcu_nocb_cb_kthread(arg = 0xFFFFFF817F245680) -008|kthread(_create = 0xFFFFFF80803122C0) -009|ret_from_fork(asm) rcuop/y -000|queued_spin_lock_slowpath(lock = 0xFFFFFFC08291BBC8, val = 0) -001|queued_spin_lock() -001|lockdep_lock() -001|graph_lock() // try to hold graph lock -002|lookup_chain_cache_add() -002|validate_chain() -003|lock_acquire -004|_raw_spin_lock_irqsave(lock = 0xFFFFFF817F211D80) -005|lock_timer_base(inline) -006|mod_timer(inline) -006|wake_nocb_gp_defer(inline)// hold nocb_gp_lock -006|__call_rcu_nocb_wake(rdp = 0xFFFFFF817F2A8680) -007|__call_rcu_common(inline) -007|call_rcu(head = 0xFFFFFFC0822E0B58, func = ?) -008|call_rcu_hurry(inline) -008|rcu_sync_call(inline) -008|rcu_sync_func(rhp = 0xFFFFFFC0822E0B58) -009|rcu_do_batch(rdp = 0xFFFFFF817F266680) -010|nocb_cb_wait(inline) -010|rcu_nocb_cb_kthread(arg = 0xFFFFFF817F266680) -011|kthread(_create = 0xFFFFFF8080363740) -012|ret_from_fork(asm) rcuop/x and rcuop/y are rcu nocb threads with the same nocb gp thread. This patch release the graph lock before lockdep call_rcu. Fixes: a0b0fd53e1e6 ("locking/lockdep: Free lock classes that are no longer in use") Cc: <stable(a)vger.kernel.org> Cc: Boqun Feng <boqun.feng(a)gmail.com> Cc: Waiman Long <longman(a)redhat.com> Cc: Carlos Llamas <cmllamas(a)google.com> Cc: Bart Van Assche <bvanassche(a)acm.org> Signed-off-by: Zhiguo Niu <zhiguo.niu(a)unisoc.com> Signed-off-by: Xuewen Yan <xuewen.yan(a)unisoc.com> Reviewed-by: Boqun Feng <boqun.feng(a)gmail.com> Reviewed-by: Waiman Long <longman(a)redhat.com> Reviewed-by: Carlos Llamas <cmllamas(a)google.com> Reviewed-by: Bart Van Assche <bvanassche(a)acm.org> Signed-off-by: Carlos Llamas <cmllamas(a)google.com> --- kernel/locking/lockdep.c | 48 ++++++++++++++++++++++++++-------------- 1 file changed, 32 insertions(+), 16 deletions(-) diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c index 151bd3de5936..3468d8230e5f 100644 --- a/kernel/locking/lockdep.c +++ b/kernel/locking/lockdep.c @@ -6184,25 +6184,27 @@ static struct pending_free *get_pending_free(void) static void free_zapped_rcu(struct rcu_head *cb); /* - * Schedule an RCU callback if no RCU callback is pending. Must be called with - * the graph lock held. - */ -static void call_rcu_zapped(struct pending_free *pf) +* See if we need to queue an RCU callback, must called with +* the lockdep lock held, returns false if either we don't have +* any pending free or the callback is already scheduled. +* Otherwise, a call_rcu() must follow this function call. +*/ +static bool prepare_call_rcu_zapped(struct pending_free *pf) { WARN_ON_ONCE(inside_selftest()); if (list_empty(&pf->zapped)) - return; + return false; if (delayed_free.scheduled) - return; + return false; delayed_free.scheduled = true; WARN_ON_ONCE(delayed_free.pf + delayed_free.index != pf); delayed_free.index ^= 1; - call_rcu(&delayed_free.rcu_head, free_zapped_rcu); + return true; } /* The caller must hold the graph lock. May be called from RCU context. */ @@ -6228,6 +6230,7 @@ static void free_zapped_rcu(struct rcu_head *ch) { struct pending_free *pf; unsigned long flags; + bool need_callback; if (WARN_ON_ONCE(ch != &delayed_free.rcu_head)) return; @@ -6239,14 +6242,18 @@ static void free_zapped_rcu(struct rcu_head *ch) pf = delayed_free.pf + (delayed_free.index ^ 1); __free_zapped_classes(pf); delayed_free.scheduled = false; + need_callback = + prepare_call_rcu_zapped(delayed_free.pf + delayed_free.index); + lockdep_unlock(); + raw_local_irq_restore(flags); /* - * If there's anything on the open list, close and start a new callback. - */ - call_rcu_zapped(delayed_free.pf + delayed_free.index); + * If there's pending free and its callback has not been scheduled, + * queue an RCU callback. + */ + if (need_callback) + call_rcu(&delayed_free.rcu_head, free_zapped_rcu); - lockdep_unlock(); - raw_local_irq_restore(flags); } /* @@ -6286,6 +6293,7 @@ static void lockdep_free_key_range_reg(void *start, unsigned long size) { struct pending_free *pf; unsigned long flags; + bool need_callback; init_data_structures_once(); @@ -6293,10 +6301,11 @@ static void lockdep_free_key_range_reg(void *start, unsigned long size) lockdep_lock(); pf = get_pending_free(); __lockdep_free_key_range(pf, start, size); - call_rcu_zapped(pf); + need_callback = prepare_call_rcu_zapped(pf); lockdep_unlock(); raw_local_irq_restore(flags); - + if (need_callback) + call_rcu(&delayed_free.rcu_head, free_zapped_rcu); /* * Wait for any possible iterators from look_up_lock_class() to pass * before continuing to free the memory they refer to. @@ -6390,6 +6399,7 @@ static void lockdep_reset_lock_reg(struct lockdep_map *lock) struct pending_free *pf; unsigned long flags; int locked; + bool need_callback = false; raw_local_irq_save(flags); locked = graph_lock(); @@ -6398,11 +6408,13 @@ static void lockdep_reset_lock_reg(struct lockdep_map *lock) pf = get_pending_free(); __lockdep_reset_lock(pf, lock); - call_rcu_zapped(pf); + need_callback = prepare_call_rcu_zapped(pf); graph_unlock(); out_irq: raw_local_irq_restore(flags); + if (need_callback) + call_rcu(&delayed_free.rcu_head, free_zapped_rcu); } /* @@ -6446,6 +6458,7 @@ void lockdep_unregister_key(struct lock_class_key *key) struct pending_free *pf; unsigned long flags; bool found = false; + bool need_callback = false; might_sleep(); @@ -6466,11 +6479,14 @@ void lockdep_unregister_key(struct lock_class_key *key) if (found) { pf = get_pending_free(); __lockdep_free_key_range(pf, key, 1); - call_rcu_zapped(pf); + need_callback = prepare_call_rcu_zapped(pf); } lockdep_unlock(); raw_local_irq_restore(flags); + if (need_callback) + call_rcu(&delayed_free.rcu_head, free_zapped_rcu); + /* Wait until is_dynamic_key() has finished accessing k->hash_entry. */ synchronize_rcu(); } -- 2.45.0.rc1.225.g2a3ae87e7f-goog

11 months

5
6
0 0

[PATCH 1/3] drm/amdgpu: Forward soft recovery errors to userspace

by Joshua Ashton

As we discussed before[1], soft recovery should be forwarded to userspace, or we can get into a really bad state where apps will keep submitting hanging command buffers cascading us to a hard reset. 1: https://lore.kernel.org/all/bf23d5ed-9a6b-43e7-84ee-8cbfd0d60f18@froggi.es/ Signed-off-by: Joshua Ashton <joshua(a)froggi.es> Cc: Friedrich Vock <friedrich.vock(a)gmx.de> Cc: Bas Nieuwenhuizen <bas(a)basnieuwenhuizen.nl> Cc: Christian König <christian.koenig(a)amd.com> Cc: André Almeida <andrealmeid(a)igalia.com> Cc: stable(a)vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 4b3000c21ef2..aebf59855e9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -262,9 +262,8 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job, struct dma_fence *fence = NULL; int r; - /* Ignore soft recovered fences here */ r = drm_sched_entity_error(s_entity); - if (r && r != -ENODATA) + if (r) goto error; if (!fence && job->gang_submit) -- 2.44.0

11 months, 1 week

5
10
0 0

[PATCH V7 1/9] sched/deadline: Comment sched_dl_entity::dl_server variable

by Daniel Bristot de Oliveira

Add an explanation for the newly added variable. Cc: stable(a)vger.kernel.org Fixes: 63ba8422f876 ("sched/deadline: Introduce deadline servers") Signed-off-by: Daniel Bristot de Oliveira <bristot(a)kernel.org> --- include/linux/sched.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/include/linux/sched.h b/include/linux/sched.h index 61591ac6eab6..abce356932cc 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -637,6 +637,8 @@ struct sched_dl_entity { * * @dl_overrun tells if the task asked to be informed about runtime * overruns. + * + * @dl_server tells if this is a server entity. */ unsigned int dl_throttled : 1; unsigned int dl_yielded : 1; -- 2.45.1

11 months, 1 week

2
1
0 0

[PATCH V7 2/9] sched/core: Add clearing of ->dl_server in put_prev_task_balance()

by Daniel Bristot de Oliveira

From: "Joel Fernandes (Google)" <joel(a)joelfernandes.org> Paths using put_prev_task_balance() need to do a pick shortly after. Make sure they also clear the ->dl_server on prev as a part of that. Cc: stable(a)vger.kernel.org Fixes: 63ba8422f876 ("sched/deadline: Introduce deadline servers") Signed-off-by: Joel Fernandes (Google) <joel(a)joelfernandes.org> Signed-off-by: Daniel Bristot de Oliveira <bristot(a)kernel.org> --- kernel/sched/core.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index bcf2c4cc0522..08c409457152 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -6006,6 +6006,14 @@ static void put_prev_task_balance(struct rq *rq, struct task_struct *prev, #endif put_prev_task(rq, prev); + + /* + * We've updated @prev and no longer need the server link, clear it. + * Must be done before ->pick_next_task() because that can (re)set + * ->dl_server. + */ + if (prev->dl_server) + prev->dl_server = NULL; } /* @@ -6049,14 +6057,6 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) restart: put_prev_task_balance(rq, prev, rf); - /* - * We've updated @prev and no longer need the server link, clear it. - * Must be done before ->pick_next_task() because that can (re)set - * ->dl_server. - */ - if (prev->dl_server) - prev->dl_server = NULL; - for_each_class(class) { p = class->pick_next_task(rq); if (p) -- 2.45.1

11 months, 1 week

2
1
0 0

[PATCH V7 3/9] sched/core: Clear prev->dl_server in CFS pick fast path

by Daniel Bristot de Oliveira

From: Youssef Esmat <youssefesmat(a)google.com> In case the previous pick was a DL server pick, ->dl_server might be set. Clear it in the fast path as well. Cc: stable(a)vger.kernel.org Fixes: 63ba8422f876 ("sched/deadline: Introduce deadline servers") Signed-off-by: Youssef Esmat <youssefesmat(a)google.com> Signed-off-by: Joel Fernandes (Google) <joel(a)joelfernandes.org> Signed-off-by: Daniel Bristot de Oliveira <bristot(a)kernel.org> --- kernel/sched/core.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 08c409457152..6d01863f93ca 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -6044,6 +6044,13 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf) p = pick_next_task_idle(rq); } + /* + * This is a normal CFS pick, but the previous could be a DL pick. + * Clear it as previous is no longer picked. + */ + if (prev->dl_server) + prev->dl_server = NULL; + /* * This is the fast path; it cannot be a DL server pick; * therefore even if @p == @prev, ->dl_server must be NULL. -- 2.45.1

11 months, 1 week

2
1
0 0

[PATCH] drm/edid: add a quirk for two 240Hz Samsung monitors

by Hamza Mahfooz

Without this fix the 5120x1440@240 timing of these monitors leads to screen flickering. Cc: stable(a)vger.kernel.org # 6.1+ Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1442 Co-developed-by: Harry Wentland <harry.wentland(a)amd.com> Signed-off-by: Harry Wentland <harry.wentland(a)amd.com> Signed-off-by: Hamza Mahfooz <hamza.mahfooz(a)amd.com> --- drivers/gpu/drm/drm_edid.c | 47 +++++++++++++++++++++++++++++++++++--- 1 file changed, 44 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index bca2af4fe1fc..3fdb8907f66b 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -89,6 +89,8 @@ static int oui(u8 first, u8 second, u8 third) #define EDID_QUIRK_NON_DESKTOP (1 << 12) /* Cap the DSC target bitrate to 15bpp */ #define EDID_QUIRK_CAP_DSC_15BPP (1 << 13) +/* Fix up a particular 5120x1440@240Hz timing */ +#define EDID_QUIRK_FIXUP_5120_1440_240 (1 << 14) #define MICROSOFT_IEEE_OUI 0xca125c @@ -170,6 +172,12 @@ static const struct edid_quirk { EDID_QUIRK('S', 'A', 'M', 596, EDID_QUIRK_PREFER_LARGE_60), EDID_QUIRK('S', 'A', 'M', 638, EDID_QUIRK_PREFER_LARGE_60), + /* Samsung C49G95T */ + EDID_QUIRK('S', 'A', 'M', 0x7053, EDID_QUIRK_FIXUP_5120_1440_240), + + /* Samsung S49AG95 */ + EDID_QUIRK('S', 'A', 'M', 0x71ac, EDID_QUIRK_FIXUP_5120_1440_240), + /* Sony PVM-2541A does up to 12 bpc, but only reports max 8 bpc */ EDID_QUIRK('S', 'N', 'Y', 0x2541, EDID_QUIRK_FORCE_12BPC), @@ -6586,7 +6594,37 @@ static void update_display_info(struct drm_connector *connector, drm_edid_to_eld(connector, drm_edid); } -static struct drm_display_mode *drm_mode_displayid_detailed(struct drm_device *dev, +static void drm_mode_displayid_detailed_edid_quirks(struct drm_connector *connector, + struct drm_display_mode *mode) +{ + unsigned int hsync_width; + unsigned int vsync_width; + + if (connector->display_info.quirks & EDID_QUIRK_FIXUP_5120_1440_240) { + if (mode->hdisplay == 5120 && mode->vdisplay == 1440 && + mode->clock == 1939490) { + hsync_width = mode->hsync_end - mode->hsync_start; + vsync_width = mode->vsync_end - mode->vsync_start; + + mode->clock = 2018490; + mode->hdisplay = 5120; + mode->hsync_start = 5120 + 8; + mode->hsync_end = 5120 + 8 + hsync_width; + mode->htotal = 5200; + + mode->vdisplay = 1440; + mode->vsync_start = 1440 + 165; + mode->vsync_end = 1440 + 165 + vsync_width; + mode->vtotal = 1619; + + drm_dbg_kms(connector->dev, + "[CONNECTOR:%d:%s] Samsung 240Hz mode quirk applied\n", + connector->base.id, connector->name); + } + } +} + +static struct drm_display_mode *drm_mode_displayid_detailed(struct drm_connector *connector, struct displayid_detailed_timings_1 *timings, bool type_7) { @@ -6605,7 +6643,7 @@ static struct drm_display_mode *drm_mode_displayid_detailed(struct drm_device *d bool hsync_positive = (timings->hsync[1] >> 7) & 0x1; bool vsync_positive = (timings->vsync[1] >> 7) & 0x1; - mode = drm_mode_create(dev); + mode = drm_mode_create(connector->dev); if (!mode) return NULL; @@ -6628,6 +6666,9 @@ static struct drm_display_mode *drm_mode_displayid_detailed(struct drm_device *d if (timings->flags & 0x80) mode->type |= DRM_MODE_TYPE_PREFERRED; + + drm_mode_displayid_detailed_edid_quirks(connector, mode); + drm_mode_set_name(mode); return mode; @@ -6650,7 +6691,7 @@ static int add_displayid_detailed_1_modes(struct drm_connector *connector, for (i = 0; i < num_timings; i++) { struct displayid_detailed_timings_1 *timings = &det->timings[i]; - newmode = drm_mode_displayid_detailed(connector->dev, timings, type_7); + newmode = drm_mode_displayid_detailed(connector, timings, type_7); if (!newmode) continue; -- 2.42.0

11 months, 3 weeks

4
5
0 0

[PATCH] usb: gadget: u_serial: Add null pointer checks after RX/TX submission

by Kuen-Han Tsai

Commit ffd603f21423 ("usb: gadget: u_serial: Add null pointer check in gs_start_io") adds null pointer checks to gs_start_io(), but it doesn't fully fix the potential null pointer dereference issue. While gserial_connect() calls gs_start_io() with port_lock held, gs_start_rx() and gs_start_tx() release the lock during endpoint request submission. This creates a window where gs_close() could set port->port_tty to NULL, leading to a dereference when the lock is reacquired. This patch adds a null pointer check for port->port_tty after RX/TX submission, and removes the initial null pointer check in gs_start_io() since the caller must hold port_lock and guarantee non-null values for port_usb and port_tty. Fixes: ffd603f21423 ("usb: gadget: u_serial: Add null pointer check in gs_start_io") Cc: stable(a)vger.kernel.org Signed-off-by: Kuen-Han Tsai <khtsai(a)google.com> --- Explanation: CPU1: CPU2: gserial_connect() // lock gs_close() // await lock gs_start_rx() // unlock usb_ep_queue() gs_close() // lock, reset port_tty and unlock gs_start_rx() // lock tty_wakeup() // dereference Stack traces: [ 51.494375][ T278] ttyGS1: shutdown [ 51.494817][ T269] android_work: sent uevent USB_STATE=DISCONNECTED [ 52.115792][ T1508] usb: [dm_bind] generic ttyGS1: super speed IN/ep1in OUT/ep1out [ 52.516288][ T1026] android_work: sent uevent USB_STATE=CONNECTED [ 52.551667][ T1533] gserial_connect: start ttyGS1 [ 52.565634][ T1533] [khtsai] enter gs_start_io, ttyGS1, port->port.tty=0000000046bd4060 [ 52.565671][ T1533] [khtsai] gs_start_rx, unlock port ttyGS1 [ 52.591552][ T1533] [khtsai] gs_start_rx, lock port ttyGS1 [ 52.619901][ T1533] [khtsai] gs_start_rx, unlock port ttyGS1 [ 52.638659][ T1325] [khtsai] gs_close, lock port ttyGS1 [ 52.656842][ T1325] gs_close: ttyGS1 (0000000046bd4060,00000000be9750a5) ... [ 52.683005][ T1325] [khtsai] gs_close, clear ttyGS1 [ 52.683007][ T1325] gs_close: ttyGS1 (0000000046bd4060,00000000be9750a5) done! [ 52.708643][ T1325] [khtsai] gs_close, unlock port ttyGS1 [ 52.747592][ T1533] [khtsai] gs_start_rx, lock port ttyGS1 [ 52.747616][ T1533] [khtsai] gs_start_io, ttyGS1, going to call tty_wakeup(), port->port.tty=0000000000000000 [ 52.747629][ T1533] Unable to handle kernel NULL pointer dereference at virtual address 00000000000001f8 --- drivers/usb/gadget/function/u_serial.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/drivers/usb/gadget/function/u_serial.c b/drivers/usb/gadget/function/u_serial.c index a92eb6d90976..2f1890c8f473 100644 --- a/drivers/usb/gadget/function/u_serial.c +++ b/drivers/usb/gadget/function/u_serial.c @@ -539,20 +539,16 @@ static int gs_alloc_requests(struct usb_ep *ep, struct list_head *head, static int gs_start_io(struct gs_port *port) { struct list_head *head = &port->read_pool; - struct usb_ep *ep; + struct usb_ep *ep = port->port_usb->out; int status; unsigned started; - if (!port->port_usb || !port->port.tty) - return -EIO; - /* Allocate RX and TX I/O buffers. We can't easily do this much * earlier (with GFP_KERNEL) because the requests are coupled to * endpoints, as are the packet sizes we'll be using. Different * configurations may use different endpoints with a given port; * and high speed vs full speed changes packet sizes too. */ - ep = port->port_usb->out; status = gs_alloc_requests(ep, head, gs_read_complete, &port->read_allocated); if (status) @@ -569,12 +565,22 @@ static int gs_start_io(struct gs_port *port) port->n_read = 0; started = gs_start_rx(port); + /* + * The TTY may be set to NULL by gs_close() after gs_start_rx() or + * gs_start_tx() release locks for endpoint request submission. + */ + if (!port->port.tty) + goto out; + if (started) { gs_start_tx(port); /* Unblock any pending writes into our circular buffer, in case * we didn't in gs_start_tx() */ + if (!port->port.tty) + goto out; tty_wakeup(port->port.tty); } else { +out: gs_free_requests(ep, head, &port->read_allocated); gs_free_requests(port->port_usb->in, &port->write_pool, &port->write_allocated); -- 2.43.0.275.g3460e3d667-goog

11 months, 3 weeks

3
7
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror May 2024