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

[PATCH V2 for 6.6] mm/migrate: correct nr_failed in migrate_pages_sync()

by Huacai Chen

From: Zi Yan <ziy(a)nvidia.com> commit a259945efe6ada94087ef666e9b38f8e34ea34ba upstream. nr_failed was missing the large folio splits from migrate_pages_batch() and can cause a mismatch between migrate_pages() return value and the number of not migrated pages, i.e., when the return value of migrate_pages() is 0, there are still pages left in the from page list. It will happen when a non-PMD THP large folio fails to migrate due to -ENOMEM and is split successfully but not all the split pages are not migrated, migrate_pages_batch() would return non-zero, but astats.nr_thp_split = 0. nr_failed would be 0 and returned to the caller of migrate_pages(), but the not migrated pages are left in the from page list without being added back to LRU lists. Fix it by adding a new nr_split counter for large folio splits and adding it to nr_failed in migrate_page_sync() after migrate_pages_batch() is done. Link: https://lkml.kernel.org/r/20231017163129.2025214-1-zi.yan@sent.com Fixes: 2ef7dbb26990 ("migrate_pages: try migrate in batch asynchronously firstly") Signed-off-by: Zi Yan <ziy(a)nvidia.com> Acked-by: Huang Ying <ying.huang(a)intel.com> Reviewed-by: Baolin Wang <baolin.wang(a)linux.alibaba.com> Cc: David Hildenbrand <david(a)redhat.com> Cc: Matthew Wilcox <willy(a)infradead.org> Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org> Signed-off-by: Huacai Chen <chenhuacai(a)loongson.cn> --- V2: Add upstream commit id. mm/migrate.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 1004b1def1c2..4ed470885217 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1504,6 +1504,7 @@ struct migrate_pages_stats { int nr_thp_succeeded; /* THP migrated successfully */ int nr_thp_failed; /* THP failed to be migrated */ int nr_thp_split; /* THP split before migrating */ + int nr_split; /* Large folio (include THP) split before migrating */ }; /* @@ -1623,6 +1624,7 @@ static int migrate_pages_batch(struct list_head *from, int nr_retry_pages = 0; int pass = 0; bool is_thp = false; + bool is_large = false; struct folio *folio, *folio2, *dst = NULL, *dst2; int rc, rc_saved = 0, nr_pages; LIST_HEAD(unmap_folios); @@ -1638,7 +1640,8 @@ static int migrate_pages_batch(struct list_head *from, nr_retry_pages = 0; list_for_each_entry_safe(folio, folio2, from, lru) { - is_thp = folio_test_large(folio) && folio_test_pmd_mappable(folio); + is_large = folio_test_large(folio); + is_thp = is_large && folio_test_pmd_mappable(folio); nr_pages = folio_nr_pages(folio); cond_resched(); @@ -1658,6 +1661,7 @@ static int migrate_pages_batch(struct list_head *from, stats->nr_thp_failed++; if (!try_split_folio(folio, split_folios)) { stats->nr_thp_split++; + stats->nr_split++; continue; } stats->nr_failed_pages += nr_pages; @@ -1686,11 +1690,12 @@ static int migrate_pages_batch(struct list_head *from, nr_failed++; stats->nr_thp_failed += is_thp; /* Large folio NUMA faulting doesn't split to retry. */ - if (folio_test_large(folio) && !nosplit) { + if (is_large && !nosplit) { int ret = try_split_folio(folio, split_folios); if (!ret) { stats->nr_thp_split += is_thp; + stats->nr_split += is_large; break; } else if (reason == MR_LONGTERM_PIN && ret == -EAGAIN) { @@ -1836,6 +1841,7 @@ static int migrate_pages_sync(struct list_head *from, new_folio_t get_new_folio, stats->nr_succeeded += astats.nr_succeeded; stats->nr_thp_succeeded += astats.nr_thp_succeeded; stats->nr_thp_split += astats.nr_thp_split; + stats->nr_split += astats.nr_split; if (rc < 0) { stats->nr_failed_pages += astats.nr_failed_pages; stats->nr_thp_failed += astats.nr_thp_failed; @@ -1843,7 +1849,11 @@ static int migrate_pages_sync(struct list_head *from, new_folio_t get_new_folio, return rc; } stats->nr_thp_failed += astats.nr_thp_split; - nr_failed += astats.nr_thp_split; + /* + * Do not count rc, as pages will be retried below. + * Count nr_split only, since it includes nr_thp_split. + */ + nr_failed += astats.nr_split; /* * Fall back to migrate all failed folios one by one synchronously. All * failed folios except split THPs will be retried, so their failure -- 2.47.1

2 months

2
1
0 0

[PATCH 6.1.y] ext4: fix uninitialized ratelimit_state->lock access in __ext4_fill_super()

by bin.lan.cn＠windriver.com

From: Baokun Li <libaokun1(a)huawei.com> [ Upstream commit b4b4fda34e535756f9e774fb2d09c4537b7dfd1c ] In the following concurrency we will access the uninitialized rs->lock: ext4_fill_super ext4_register_sysfs // sysfs registered msg_ratelimit_interval_ms // Other processes modify rs->interval to // non-zero via msg_ratelimit_interval_ms ext4_orphan_cleanup ext4_msg(sb, KERN_INFO, "Errors on filesystem, " __ext4_msg ___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state) if (!rs->interval) // do nothing if interval is 0 return 1; raw_spin_trylock_irqsave(&rs->lock, flags) raw_spin_trylock(lock) _raw_spin_trylock __raw_spin_trylock spin_acquire(&lock->dep_map, 0, 1, _RET_IP_) lock_acquire __lock_acquire register_lock_class assign_lock_key dump_stack(); ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); raw_spin_lock_init(&rs->lock); // init rs->lock here and get the following dump_stack: ========================================================= INFO: trying to register non-static key. The code is fine but needs lockdep annotation, or maybe you didn't initialize this object before use? turning off the locking correctness validator. CPU: 12 PID: 753 Comm: mount Tainted: G E 6.7.0-rc6-next-20231222 #504 [...] Call Trace: dump_stack_lvl+0xc5/0x170 dump_stack+0x18/0x30 register_lock_class+0x740/0x7c0 __lock_acquire+0x69/0x13a0 lock_acquire+0x120/0x450 _raw_spin_trylock+0x98/0xd0 ___ratelimit+0xf6/0x220 __ext4_msg+0x7f/0x160 [ext4] ext4_orphan_cleanup+0x665/0x740 [ext4] __ext4_fill_super+0x21ea/0x2b10 [ext4] ext4_fill_super+0x14d/0x360 [ext4] [...] ========================================================= Normally interval is 0 until s_msg_ratelimit_state is initialized, so ___ratelimit() does nothing. But registering sysfs precedes initializing rs->lock, so it is possible to change rs->interval to a non-zero value via the msg_ratelimit_interval_ms interface of sysfs while rs->lock is uninitialized, and then a call to ext4_msg triggers the problem by accessing an uninitialized rs->lock. Therefore register sysfs after all initializations are complete to avoid such problems. Signed-off-by: Baokun Li <libaokun1(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Link: https://lore.kernel.org/r/20240102133730.1098120-1-libaokun1@huawei.com Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> [Minor context change fixed.] Signed-off-by: Bin Lan <bin.lan.cn(a)windriver.com> Signed-off-by: He Zhe <zhe.he(a)windriver.com> --- Build test passed. --- fs/ext4/super.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-) diff --git a/fs/ext4/super.c b/fs/ext4/super.c index 7f0231b34905..8528f61854ab 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -5496,19 +5496,15 @@ static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) if (err) goto failed_mount6; - err = ext4_register_sysfs(sb); - if (err) - goto failed_mount7; - err = ext4_init_orphan_info(sb); if (err) - goto failed_mount8; + goto failed_mount7; #ifdef CONFIG_QUOTA /* Enable quota usage during mount. */ if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { err = ext4_enable_quotas(sb); if (err) - goto failed_mount9; + goto failed_mount8; } #endif /* CONFIG_QUOTA */ @@ -5534,7 +5530,7 @@ static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) ext4_msg(sb, KERN_INFO, "recovery complete"); err = ext4_mark_recovery_complete(sb, es); if (err) - goto failed_mount10; + goto failed_mount9; } if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev)) @@ -5551,15 +5547,17 @@ static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) atomic_set(&sbi->s_warning_count, 0); atomic_set(&sbi->s_msg_count, 0); + /* Register sysfs after all initializations are complete. */ + err = ext4_register_sysfs(sb); + if (err) + goto failed_mount9; + return 0; -failed_mount10: +failed_mount9: ext4_quota_off_umount(sb); -failed_mount9: __maybe_unused +failed_mount8: __maybe_unused ext4_release_orphan_info(sb); -failed_mount8: - ext4_unregister_sysfs(sb); - kobject_put(&sbi->s_kobj); failed_mount7: ext4_unregister_li_request(sb); failed_mount6: -- 2.34.1

2 months

2
1
0 0

FAILED: patch "[PATCH] clocksource/i8253: Use raw_spinlock_irqsave() in" failed to apply to 5.4-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.4-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.4.y git checkout FETCH_HEAD git cherry-pick -x 94cff94634e506a4a44684bee1875d2dbf782722 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025051258-washbowl-alongside-de3d@gregkh' --subject-prefix 'PATCH 5.4.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 94cff94634e506a4a44684bee1875d2dbf782722 Mon Sep 17 00:00:00 2001 From: Sebastian Andrzej Siewior <bigeasy(a)linutronix.de> Date: Fri, 4 Apr 2025 15:31:16 +0200 Subject: [PATCH] clocksource/i8253: Use raw_spinlock_irqsave() in clockevent_i8253_disable() On x86 during boot, clockevent_i8253_disable() can be invoked via x86_late_time_init -> hpet_time_init() -> pit_timer_init() which happens with enabled interrupts. If some of the old i8253 hardware is actually used then lockdep will notice that i8253_lock is used in hard interrupt context. This causes lockdep to complain because it observed the lock being acquired with interrupts enabled and in hard interrupt context. Make clockevent_i8253_disable() acquire the lock with raw_spinlock_irqsave() to cure this. [ tglx: Massage change log and use guard() ] Fixes: c8c4076723dac ("x86/timer: Skip PIT initialization on modern chipsets") Signed-off-by: Sebastian Andrzej Siewior <bigeasy(a)linutronix.de> Signed-off-by: Thomas Gleixner <tglx(a)linutronix.de> Cc: stable(a)vger.kernel.org Link: https://lore.kernel.org/all/20250404133116.p-XRWJXf@linutronix.de diff --git a/drivers/clocksource/i8253.c b/drivers/clocksource/i8253.c index 39f7c2d736d1..b603c25f3dfa 100644 --- a/drivers/clocksource/i8253.c +++ b/drivers/clocksource/i8253.c @@ -103,7 +103,7 @@ int __init clocksource_i8253_init(void) #ifdef CONFIG_CLKEVT_I8253 void clockevent_i8253_disable(void) { - raw_spin_lock(&i8253_lock); + guard(raw_spinlock_irqsave)(&i8253_lock); /* * Writing the MODE register should stop the counter, according to @@ -132,8 +132,6 @@ void clockevent_i8253_disable(void) outb_p(0, PIT_CH0); outb_p(0x30, PIT_MODE); - - raw_spin_unlock(&i8253_lock); } static int pit_shutdown(struct clock_event_device *evt)

2 months

3
2
0 0

FAILED: patch "[PATCH] selftests/mm: compaction_test: support platform with huge" failed to apply to 5.10-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.10-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.10.y git checkout FETCH_HEAD git cherry-pick -x ab00ddd802f80e31fc9639c652d736fe3913feae # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025051245-jailbreak-unlinked-27ec@gregkh' --subject-prefix 'PATCH 5.10.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From ab00ddd802f80e31fc9639c652d736fe3913feae Mon Sep 17 00:00:00 2001 From: Feng Tang <feng.tang(a)linux.alibaba.com> Date: Wed, 23 Apr 2025 18:36:45 +0800 Subject: [PATCH] selftests/mm: compaction_test: support platform with huge mount of memory When running mm selftest to verify mm patches, 'compaction_test' case failed on an x86 server with 1TB memory. And the root cause is that it has too much free memory than what the test supports. The test case tries to allocate 100000 huge pages, which is about 200 GB for that x86 server, and when it succeeds, it expects it's large than 1/3 of 80% of the free memory in system. This logic only works for platform with 750 GB ( 200 / (1/3) / 80% ) or less free memory, and may raise false alarm for others. Fix it by changing the fixed page number to self-adjustable number according to the real number of free memory. Link: https://lkml.kernel.org/r/20250423103645.2758-1-feng.tang@linux.alibaba.com Fixes: bd67d5c15cc1 ("Test compaction of mlocked memory") Signed-off-by: Feng Tang <feng.tang(a)linux.alibaba.com> Acked-by: Dev Jain <dev.jain(a)arm.com> Reviewed-by: Baolin Wang <baolin.wang(a)linux.alibaba.com> Tested-by: Baolin Wang <baolin.wang(a)inux.alibaba.com> Cc: Shuah Khan <shuah(a)kernel.org> Cc: Sri Jayaramappa <sjayaram(a)akamai.com> Cc: <stable(a)vger.kernel.org> Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org> diff --git a/tools/testing/selftests/mm/compaction_test.c b/tools/testing/selftests/mm/compaction_test.c index 2c3a0eb6b22d..9bc4591c7b16 100644 --- a/tools/testing/selftests/mm/compaction_test.c +++ b/tools/testing/selftests/mm/compaction_test.c @@ -90,6 +90,8 @@ int check_compaction(unsigned long mem_free, unsigned long hugepage_size, int compaction_index = 0; char nr_hugepages[20] = {0}; char init_nr_hugepages[24] = {0}; + char target_nr_hugepages[24] = {0}; + int slen; snprintf(init_nr_hugepages, sizeof(init_nr_hugepages), "%lu", initial_nr_hugepages); @@ -106,11 +108,18 @@ int check_compaction(unsigned long mem_free, unsigned long hugepage_size, goto out; } - /* Request a large number of huge pages. The Kernel will allocate - as much as it can */ - if (write(fd, "100000", (6*sizeof(char))) != (6*sizeof(char))) { - ksft_print_msg("Failed to write 100000 to /proc/sys/vm/nr_hugepages: %s\n", - strerror(errno)); + /* + * Request huge pages for about half of the free memory. The Kernel + * will allocate as much as it can, and we expect it will get at least 1/3 + */ + nr_hugepages_ul = mem_free / hugepage_size / 2; + snprintf(target_nr_hugepages, sizeof(target_nr_hugepages), + "%lu", nr_hugepages_ul); + + slen = strlen(target_nr_hugepages); + if (write(fd, target_nr_hugepages, slen) != slen) { + ksft_print_msg("Failed to write %lu to /proc/sys/vm/nr_hugepages: %s\n", + nr_hugepages_ul, strerror(errno)); goto close_fd; }

2 months

3
2
0 0

[PATCH V2 1/2] drm/amdgpu: read back register after written for VCN v4.0.5

by David (Ming Qiang) Wu

V2: not to add extra read-back in vcn_v4_0_5_start as there is a read-back already. New comment for better understanding. On VCN v4.0.5 there is a race condition where the WPTR is not updated after starting from idle when doorbell is used. The read-back of regVCN_RB1_DB_CTRL register after written is to ensure the doorbell_index is updated before it can work properly. Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528 Cc: stable(a)vger.kernel.org Signed-off-by: David (Ming Qiang) Wu <David.Wu3(a)amd.com> Reviewed-by: Mario Limonciello <mario.limonciello(a)amd.com> Tested-by: Mario Limonciello <mario.limonciello(a)amd.com> --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c index ed00d35039c13..e55b76d71367d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c @@ -1034,6 +1034,10 @@ static int vcn_v4_0_5_start_dpg_mode(struct amdgpu_vcn_inst *vinst, ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | VCN_RB1_DB_CTRL__EN_MASK); + /* Keeping one read-back to ensure all register writes are done, otherwise + * it may introduce race conditions */ + RREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL); + return 0; } -- 2.34.1

2 months

2
1
0 0

[PATCH v5] x86/sev: Fix making shared pages private during kdump

by Ashish Kalra

From: Ashish Kalra <ashish.kalra(a)amd.com> When the shared pages are being made private during kdump preparation there are additional checks to handle shared GHCB pages. These additional checks include handling the case of GHCB page being contained within a huge page. The check for handling the case of GHCB contained within a huge page incorrectly skips a page just below the GHCB page from being transitioned back to private during kdump preparation. This skipped page causes a 0x404 #VC exception when it is accessed later while dumping guest memory during vmcore generation via kdump. Correct the range to be checked for GHCB contained in a huge page. Also ensure that the skipped huge page containing the GHCB page is transitioned back to private by applying the correct address mask later when changing GHCBs to private at end of kdump preparation. Fixes: 3074152e56c9 ("x86/sev: Convert shared memory back to private on kexec") Cc: stable(a)vger.kernel.org Signed-off-by: Ashish Kalra <ashish.kalra(a)amd.com> --- arch/x86/coco/sev/core.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/arch/x86/coco/sev/core.c b/arch/x86/coco/sev/core.c index d35fec7b164a..30b74e4e4e88 100644 --- a/arch/x86/coco/sev/core.c +++ b/arch/x86/coco/sev/core.c @@ -1019,7 +1019,8 @@ static void unshare_all_memory(void) data = per_cpu(runtime_data, cpu); ghcb = (unsigned long)&data->ghcb_page; - if (addr <= ghcb && ghcb <= addr + size) { + /* Handle the case of a huge page containing the GHCB page */ + if (addr <= ghcb && ghcb < addr + size) { skipped_addr = true; break; } @@ -1131,8 +1132,8 @@ static void shutdown_all_aps(void) void snp_kexec_finish(void) { struct sev_es_runtime_data *data; + unsigned long size, addr; unsigned int level, cpu; - unsigned long size; struct ghcb *ghcb; pte_t *pte; @@ -1160,8 +1161,10 @@ void snp_kexec_finish(void) ghcb = &data->ghcb_page; pte = lookup_address((unsigned long)ghcb, &level); size = page_level_size(level); - set_pte_enc(pte, level, (void *)ghcb); - snp_set_memory_private((unsigned long)ghcb, (size / PAGE_SIZE)); + /* Handle the case of a huge page containing the GHCB page */ + addr = (unsigned long)ghcb & page_level_mask(level); + set_pte_enc(pte, level, (void *)addr); + snp_set_memory_private(addr, (size / PAGE_SIZE)); } } -- 2.34.1

2 months

5
4
0 0

[PATCH 1/2] drm/amdgpu: read back DB_CTRL register after written for VCN v4.0.5

by David (Ming Qiang) Wu

On VCN v4.0.5 there is a race condition where the WPTR is not updated after starting from idle when doorbell is used. The read-back of regVCN_RB1_DB_CTRL register after written is to ensure the doorbell_index is updated before it can work properly. Link: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528 Signed-off-by: David (Ming Qiang) Wu <David.Wu3(a)amd.com> --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c index ed00d35039c1..d6be8b05d7a2 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c @@ -1033,6 +1033,8 @@ static int vcn_v4_0_5_start_dpg_mode(struct amdgpu_vcn_inst *vinst, WREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL, ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | VCN_RB1_DB_CTRL__EN_MASK); + /* Read DB_CTRL to flush the write DB_CTRL command. */ + RREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL); return 0; } @@ -1195,6 +1197,8 @@ static int vcn_v4_0_5_start(struct amdgpu_vcn_inst *vinst) WREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL, ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | VCN_RB1_DB_CTRL__EN_MASK); + /* Read DB_CTRL to flush the write DB_CTRL command. */ + RREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL); WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, ring->gpu_addr); WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); -- 2.49.0

2 months

5
8
0 0

[PATCH 1/3] perf/arm-ni: Set initial IRQ affinity

by Robin Murphy

While we do request our IRQs with the right flags to stop their affinity changing unexpectedly, we forgot to actually set it to start with. Oops. Cc: stable(a)vger.kernel.org Fixes: 4d5a7680f2b4 ("perf: Add driver for Arm NI-700 interconnect PMU") Signed-off-by: Robin Murphy <robin.murphy(a)arm.com> --- drivers/perf/arm-ni.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/perf/arm-ni.c b/drivers/perf/arm-ni.c index de7b6cce4d68..9396d243415f 100644 --- a/drivers/perf/arm-ni.c +++ b/drivers/perf/arm-ni.c @@ -544,6 +544,8 @@ static int arm_ni_init_cd(struct arm_ni *ni, struct arm_ni_node *node, u64 res_s return err; cd->cpu = cpumask_local_spread(0, dev_to_node(ni->dev)); + irq_set_affinity(cd->irq, cpumask_of(cd->cpu)); + cd->pmu = (struct pmu) { .module = THIS_MODULE, .parent = ni->dev, -- 2.39.2.101.g768bb238c484.dirty

2 months

1
0
0 0

[PATCH v2 3/3] can: kvaser_pciefd: Continue parsing DMA buf after dropped RX

by Axel Forsman

Going bus-off on a channel doing RX could result in dropped packets. As netif_running() gets cleared before the channel abort procedure, the handling of any last RDATA packets would see netif_rx() return non-zero to signal a dropped packet. kvaser_pciefd_read_buffer() dealt with this "error" by breaking out of processing the remaining DMA RX buffer. Only return an error from kvaser_pciefd_read_buffer() due to packet corruption, otherwise handle it internally. Cc: stable(a)vger.kernel.org Signed-off-by: Axel Forsman <axfo(a)kvaser.com> Tested-by: Jimmy Assarsson <extja(a)kvaser.com> Reviewed-by: Jimmy Assarsson <extja(a)kvaser.com> --- drivers/net/can/kvaser_pciefd.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/drivers/net/can/kvaser_pciefd.c b/drivers/net/can/kvaser_pciefd.c index 6cb657018cdf..6b3b04b5a4c2 100644 --- a/drivers/net/can/kvaser_pciefd.c +++ b/drivers/net/can/kvaser_pciefd.c @@ -1209,7 +1209,7 @@ static int kvaser_pciefd_handle_data_packet(struct kvaser_pciefd *pcie, skb = alloc_canfd_skb(priv->dev, &cf); if (!skb) { priv->dev->stats.rx_dropped++; - return -ENOMEM; + return 0; } cf->len = can_fd_dlc2len(dlc); @@ -1221,7 +1221,7 @@ static int kvaser_pciefd_handle_data_packet(struct kvaser_pciefd *pcie, skb = alloc_can_skb(priv->dev, (struct can_frame **)&cf); if (!skb) { priv->dev->stats.rx_dropped++; - return -ENOMEM; + return 0; } can_frame_set_cc_len((struct can_frame *)cf, dlc, priv->ctrlmode); } @@ -1239,7 +1239,9 @@ static int kvaser_pciefd_handle_data_packet(struct kvaser_pciefd *pcie, priv->dev->stats.rx_packets++; kvaser_pciefd_set_skb_timestamp(pcie, skb, p->timestamp); - return netif_rx(skb); + netif_rx(skb); + + return 0; } static void kvaser_pciefd_change_state(struct kvaser_pciefd_can *can, -- 2.47.2

2 months

1
0
0 0

[PATCH v2 2/3] can: kvaser_pciefd: Fix echo_skb race conditions

by Axel Forsman

The functions kvaser_pciefd_start_xmit() and kvaser_pciefd_handle_ack_packet() raced to stop/wake TX queues and get/put echo skbs, as kvaser_pciefd_can->echo_lock was only ever taken when transmitting. E.g., this caused the following error: can_put_echo_skb: BUG! echo_skb 5 is occupied! Instead, use the synchronization helpers in netdev_queues.h. As those piggyback on BQL barriers, start updating in-flight packets and bytes counts as well. Cc: stable(a)vger.kernel.org Signed-off-by: Axel Forsman <axfo(a)kvaser.com> Tested-by: Jimmy Assarsson <extja(a)kvaser.com> Reviewed-by: Jimmy Assarsson <extja(a)kvaser.com> --- drivers/net/can/kvaser_pciefd.c | 91 +++++++++++++++++++++------------ 1 file changed, 57 insertions(+), 34 deletions(-) diff --git a/drivers/net/can/kvaser_pciefd.c b/drivers/net/can/kvaser_pciefd.c index 4d84f7d13c6f..6cb657018cdf 100644 --- a/drivers/net/can/kvaser_pciefd.c +++ b/drivers/net/can/kvaser_pciefd.c @@ -16,6 +16,7 @@ #include <linux/netdevice.h> #include <linux/pci.h> #include <linux/timer.h> +#include <net/netdev_queues.h> MODULE_LICENSE("Dual BSD/GPL"); MODULE_AUTHOR("Kvaser AB <support(a)kvaser.com>"); @@ -410,10 +411,13 @@ struct kvaser_pciefd_can { void __iomem *reg_base; struct can_berr_counter bec; u8 cmd_seq; + u8 tx_max_count; + u8 tx_idx; + u8 ack_idx; int err_rep_cnt; - int echo_idx; + unsigned int completed_tx_pkts; + unsigned int completed_tx_bytes; spinlock_t lock; /* Locks sensitive registers (e.g. MODE) */ - spinlock_t echo_lock; /* Locks the message echo buffer */ struct timer_list bec_poll_timer; struct completion start_comp, flush_comp; }; @@ -714,6 +718,9 @@ static int kvaser_pciefd_open(struct net_device *netdev) int ret; struct kvaser_pciefd_can *can = netdev_priv(netdev); + can->tx_idx = 0; + can->ack_idx = 0; + ret = open_candev(netdev); if (ret) return ret; @@ -745,21 +752,26 @@ static int kvaser_pciefd_stop(struct net_device *netdev) del_timer(&can->bec_poll_timer); } can->can.state = CAN_STATE_STOPPED; + netdev_reset_queue(netdev); close_candev(netdev); return ret; } +static unsigned int kvaser_pciefd_tx_avail(const struct kvaser_pciefd_can *can) +{ + return can->tx_max_count - (READ_ONCE(can->tx_idx) - READ_ONCE(can->ack_idx)); +} + static int kvaser_pciefd_prepare_tx_packet(struct kvaser_pciefd_tx_packet *p, - struct kvaser_pciefd_can *can, + struct can_priv *can, u8 seq, struct sk_buff *skb) { struct canfd_frame *cf = (struct canfd_frame *)skb->data; int packet_size; - int seq = can->echo_idx; memset(p, 0, sizeof(*p)); - if (can->can.ctrlmode & CAN_CTRLMODE_ONE_SHOT) + if (can->ctrlmode & CAN_CTRLMODE_ONE_SHOT) p->header[1] |= KVASER_PCIEFD_TPACKET_SMS; if (cf->can_id & CAN_RTR_FLAG) @@ -782,7 +794,7 @@ static int kvaser_pciefd_prepare_tx_packet(struct kvaser_pciefd_tx_packet *p, } else { p->header[1] |= FIELD_PREP(KVASER_PCIEFD_RPACKET_DLC_MASK, - can_get_cc_dlc((struct can_frame *)cf, can->can.ctrlmode)); + can_get_cc_dlc((struct can_frame *)cf, can->ctrlmode)); } p->header[1] |= FIELD_PREP(KVASER_PCIEFD_PACKET_SEQ_MASK, seq); @@ -797,22 +809,24 @@ static netdev_tx_t kvaser_pciefd_start_xmit(struct sk_buff *skb, struct net_device *netdev) { struct kvaser_pciefd_can *can = netdev_priv(netdev); - unsigned long irq_flags; struct kvaser_pciefd_tx_packet packet; + unsigned int seq = can->tx_idx & (can->can.echo_skb_max - 1); + unsigned int frame_len; int nr_words; - u8 count; if (can_dev_dropped_skb(netdev, skb)) return NETDEV_TX_OK; + if (!netif_subqueue_maybe_stop(netdev, 0, kvaser_pciefd_tx_avail(can), 1, 1)) + return NETDEV_TX_BUSY; - nr_words = kvaser_pciefd_prepare_tx_packet(&packet, can, skb); + nr_words = kvaser_pciefd_prepare_tx_packet(&packet, &can->can, seq, skb); - spin_lock_irqsave(&can->echo_lock, irq_flags); /* Prepare and save echo skb in internal slot */ - can_put_echo_skb(skb, netdev, can->echo_idx, 0); - - /* Move echo index to the next slot */ - can->echo_idx = (can->echo_idx + 1) % can->can.echo_skb_max; + WRITE_ONCE(can->can.echo_skb[seq], NULL); + frame_len = can_skb_get_frame_len(skb); + can_put_echo_skb(skb, netdev, seq, frame_len); + netdev_sent_queue(netdev, frame_len); + WRITE_ONCE(can->tx_idx, can->tx_idx + 1); /* Write header to fifo */ iowrite32(packet.header[0], @@ -836,14 +850,7 @@ static netdev_tx_t kvaser_pciefd_start_xmit(struct sk_buff *skb, KVASER_PCIEFD_KCAN_FIFO_LAST_REG); } - count = FIELD_GET(KVASER_PCIEFD_KCAN_TX_NR_PACKETS_CURRENT_MASK, - ioread32(can->reg_base + KVASER_PCIEFD_KCAN_TX_NR_PACKETS_REG)); - /* No room for a new message, stop the queue until at least one - * successful transmit - */ - if (count >= can->can.echo_skb_max || can->can.echo_skb[can->echo_idx]) - netif_stop_queue(netdev); - spin_unlock_irqrestore(&can->echo_lock, irq_flags); + netif_subqueue_maybe_stop(netdev, 0, kvaser_pciefd_tx_avail(can), 1, 1); return NETDEV_TX_OK; } @@ -970,6 +977,8 @@ static int kvaser_pciefd_setup_can_ctrls(struct kvaser_pciefd *pcie) can->kv_pcie = pcie; can->cmd_seq = 0; can->err_rep_cnt = 0; + can->completed_tx_pkts = 0; + can->completed_tx_bytes = 0; can->bec.txerr = 0; can->bec.rxerr = 0; @@ -983,11 +992,10 @@ static int kvaser_pciefd_setup_can_ctrls(struct kvaser_pciefd *pcie) tx_nr_packets_max = FIELD_GET(KVASER_PCIEFD_KCAN_TX_NR_PACKETS_MAX_MASK, ioread32(can->reg_base + KVASER_PCIEFD_KCAN_TX_NR_PACKETS_REG)); + can->tx_max_count = min(KVASER_PCIEFD_CAN_TX_MAX_COUNT, tx_nr_packets_max - 1); can->can.clock.freq = pcie->freq; - can->can.echo_skb_max = min(KVASER_PCIEFD_CAN_TX_MAX_COUNT, tx_nr_packets_max - 1); - can->echo_idx = 0; - spin_lock_init(&can->echo_lock); + can->can.echo_skb_max = roundup_pow_of_two(can->tx_max_count); spin_lock_init(&can->lock); can->can.bittiming_const = &kvaser_pciefd_bittiming_const; @@ -1510,19 +1518,21 @@ static int kvaser_pciefd_handle_ack_packet(struct kvaser_pciefd *pcie, netdev_dbg(can->can.dev, "Packet was flushed\n"); } else { int echo_idx = FIELD_GET(KVASER_PCIEFD_PACKET_SEQ_MASK, p->header[0]); - int len; - u8 count; + unsigned int len, frame_len = 0; struct sk_buff *skb; + if (echo_idx != (can->ack_idx & (can->can.echo_skb_max - 1))) + return 0; skb = can->can.echo_skb[echo_idx]; - if (skb) - kvaser_pciefd_set_skb_timestamp(pcie, skb, p->timestamp); - len = can_get_echo_skb(can->can.dev, echo_idx, NULL); - count = FIELD_GET(KVASER_PCIEFD_KCAN_TX_NR_PACKETS_CURRENT_MASK, - ioread32(can->reg_base + KVASER_PCIEFD_KCAN_TX_NR_PACKETS_REG)); + if (!skb) + return 0; + kvaser_pciefd_set_skb_timestamp(pcie, skb, p->timestamp); + len = can_get_echo_skb(can->can.dev, echo_idx, &frame_len); - if (count < can->can.echo_skb_max && netif_queue_stopped(can->can.dev)) - netif_wake_queue(can->can.dev); + /* Pairs with barrier in kvaser_pciefd_start_xmit() */ + smp_store_release(&can->ack_idx, can->ack_idx + 1); + can->completed_tx_pkts++; + can->completed_tx_bytes += frame_len; if (!one_shot_fail) { can->can.dev->stats.tx_bytes += len; @@ -1638,11 +1646,26 @@ static int kvaser_pciefd_read_buffer(struct kvaser_pciefd *pcie, int dma_buf) { int pos = 0; int res = 0; + unsigned int i; do { res = kvaser_pciefd_read_packet(pcie, &pos, dma_buf); } while (!res && pos > 0 && pos < KVASER_PCIEFD_DMA_SIZE); + /* Report ACKs in this buffer to BQL en masse for correct periods */ + for (i = 0; i < pcie->nr_channels; ++i) { + struct kvaser_pciefd_can *can = pcie->can[i]; + + if (!can->completed_tx_pkts) + continue; + netif_subqueue_completed_wake(can->can.dev, 0, + can->completed_tx_pkts, + can->completed_tx_bytes, + kvaser_pciefd_tx_avail(can), 1); + can->completed_tx_pkts = 0; + can->completed_tx_bytes = 0; + } + return res; } -- 2.47.2

2 months

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror