November 2023 - Kernel - mailweb.openeuler.org

[PATCH openEuler-22.03-LTS 0/4] Add script to check & update openeuler_defconfig
by Jialin Zhang 25 Nov '23

25 Nov '23

Jialin Zhang (3): Revert "kconfig: Add script to update openeuler_defconfig" config: update openeuler_defconfig for x86 config: update openeuler_defconfig for arm64 Xie XiuQi (1): kconfig: Add script to check & update openeuler_defconfig arch/arm64/configs/openeuler_defconfig | 32 ++++++++++++++++++-------- arch/x86/configs/openeuler_defconfig | 12 ++++++---- scripts/kconfig/Makefile | 14 +++++++++++ scripts/kconfig/makeconfig.sh | 5 ++++ 4 files changed, 49 insertions(+), 14 deletions(-) -- 2.25.1

2 5

[PATCH openEuler-22.03-LTS-SP1 0/4] Add script to check & update openeuler_defconfig
by Jialin Zhang 25 Nov '23

25 Nov '23

Jialin Zhang (3): Revert "kconfig: Add script to update openeuler_defconfig" config: update openeuler_defconfig for x86 config: update openeuler_defconfig for arm64 Xie XiuQi (1): kconfig: Add script to check & update openeuler_defconfig arch/arm64/configs/openeuler_defconfig | 26 ++++++++++++++++++-------- arch/x86/configs/openeuler_defconfig | 8 +++----- scripts/kconfig/Makefile | 14 ++++++++++++++ scripts/kconfig/makeconfig.sh | 5 +++++ 4 files changed, 40 insertions(+), 13 deletions(-) -- 2.25.1

2 5

[PATCH openEuler-22.03-LTS-SP2 0/4] Add script to check & update openeuler_defconfig
by Jialin Zhang 25 Nov '23

25 Nov '23

Jialin Zhang (3): Revert "kconfig: Add script to update openeuler_defconfig" config: update openeuler_defconfig for x86 config: update openeuler_defconfig for arm64 Xie XiuQi (1): kconfig: Add script to check & update openeuler_defconfig arch/arm64/configs/openeuler_defconfig | 51 +++++++++++++++++--------- arch/x86/configs/openeuler_defconfig | 32 ++++++++++------ scripts/kconfig/Makefile | 14 +++++++ scripts/kconfig/makeconfig.sh | 5 +++ 4 files changed, 74 insertions(+), 28 deletions(-) -- 2.25.1

2 5

[PATCH openEuler-1.0-LTS] x86/mce/inject: Fix a wrong assignment of i_mce.status
by liwei 25 Nov '23

25 Nov '23

From: Zhenzhong Duan <zhenzhong.duan(a)gmail.com> mainline inclusion from mainline-v5.9-rc1 commit 5d7f7d1d5e01c22894dee7c9c9266500478dca99 category: bugfix bugzilla: 189383 CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… --------------------------- The original code is a nop as i_mce.status is or'ed with part of itself, fix it. Fixes: a1300e505297 ("x86/ras/mce_amd_inj: Trigger deferred and thresholding errors interrupts") Signed-off-by: Zhenzhong Duan <zhenzhong.duan(a)gmail.com> Signed-off-by: Borislav Petkov <bp(a)suse.de> Acked-by: Yazen Ghannam <yazen.ghannam(a)amd.com> Link: https://lkml.kernel.org/r/20200611023238.3830-1-zhenzhong.duan@gmail.com Signed-off-by: liwei <liwei728(a)huawei.com> --- arch/x86/kernel/cpu/mce/inject.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/kernel/cpu/mce/inject.c b/arch/x86/kernel/cpu/mce/inject.c index 3f82afd0f46f..f21f3f26c555 100644 --- a/arch/x86/kernel/cpu/mce/inject.c +++ b/arch/x86/kernel/cpu/mce/inject.c @@ -518,7 +518,7 @@ static void do_inject(void) */ if (inj_type == DFR_INT_INJ) { i_mce.status |= MCI_STATUS_DEFERRED; - i_mce.status |= (i_mce.status & ~MCI_STATUS_UC); + i_mce.status &= ~MCI_STATUS_UC; } /* -- 2.25.1

2 1

[PATCH OLK-6.6] io_uring: fix soft lockup in io_submit_sqes()
by Zizhi Wo 25 Nov '23

25 Nov '23

From: Guo Xuenan <guoxuenan(a)huawei.com> hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5930C ------------------- when set up sq ring size with IORING_MAX_ENTRIES, io_submit_sqes may looping ~32768 times which may trigger soft lockups. add cond_resched condition to avoid this bad situation. set sq ring size 32768 to perform stress test as follows: watchdog: BUG: soft lockup - CPU#3 stuck for 26s! [poc:691] Modules linked in: CPU: 3 PID: 691 Comm: poc Not tainted 5.18.0+ #9 Hardware name: linux,dummy-virt (DT) pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : arch_local_irq_enable+0xc/0x28 lr : io_issue_sqe+0x870/0x28e8 sp : ffff80000e0f7800 x29: ffff80000e0f7800 x28: ffff0000cf850dd0 x27: ffff0000cf2c2000 x26: ffff8000096f4b20 x25: ffff0000cd769c00 x24: ffff0000c12b6650 x23: ffff800009dad958 x22: 00000006fc23ac00 x21: ffff0000cd769c08 x20: 1ffff00001c1ef1a x19: ffff0000cd767e00 x18: 0000000000000000 x17: ffff800008032b74 x16: ffff800008636448 x15: 0000fffff7166568 x14: ffff80000861edf0 x13: ffff600019e58449 x12: 1fffe00019e58448 x11: 1fffe00019e58448 x10: ffff600019e58448 x9 : dfff800000000000 x8 : ffff0000cf2c2244 x7 : 0000000000000001 x6 : ffff600019e58449 x5 : ffff600019e58449 x4 : ffff600019e58449 x3 : ffff8000086306c0 x2 : 0000000000000001 x1 : ffff0000cf2c2244 x0 : 00000000000000e0 Call trace: arch_local_irq_enable+0xc/0x28 io_submit_sqes+0x530/0x29d8 __arm64_sys_io_uring_enter+0x380/0xd18 invoke_syscall+0x64/0x180 el0_svc_common.constprop.2+0x178/0x208 do_el0_svc+0x84/0xa0 el0_svc+0x48/0x1a0 el0t_64_sync_handler+0x90/0xb8 el0t_64_sync+0x180/0x184 Kernel panic - not syncing: softlockup: hung tasks CPU: 3 PID: 691 Comm: poc Tainted: G L 5.18.0+ #9 Hardware name: linux,dummy-virt (DT) Call trace: dump_backtrace+0x218/0x228 show_stack+0x20/0x68 dump_stack_lvl+0x68/0x84 dump_stack+0x1c/0x38 panic+0x1ec/0x3ec watchdog_timer_fn+0x28c/0x300 __hrtimer_run_queues+0x1d8/0x498 hrtimer_interrupt+0x238/0x558 arch_timer_handler_virt+0x48/0x60 handle_percpu_devid_irq+0xdc/0x270 generic_handle_domain_irq+0x50/0x70 gic_handle_irq+0x8c/0x4bc call_on_irq_stack+0x2c/0x38 do_interrupt_handler+0xc4/0xc8 el1_interrupt+0x48/0xb0 el1h_64_irq_handler+0x18/0x28 el1h_64_irq+0x74/0x78 arch_local_irq_enable+0xc/0x28 io_submit_sqes+0x530/0x29d8 __arm64_sys_io_uring_enter+0x380/0xd18 invoke_syscall+0x64/0x180 el0_svc_common.constprop.2+0x178/0x208 do_el0_svc+0x84/0xa0 el0_svc+0x48/0x1a0 el0t_64_sync_handler+0x90/0xb8 el0t_64_sync+0x180/0x184 SMP: stopping secondary CPUs Kernel Offset: disabled CPU features: 0x110,00008f09,00001006 Memory Limit: none ---[ end Kernel panic - not syncing: softlockup: hung tasks ]--- Link: https://lore.kernel.org/all/d4bc3afb-02d5-1793-cffa-e15b2bdb0028@huawei.com/ Signed-off-by: Guo Xuenan <guoxuenan(a)huawei.com> Conflict: io_uring/io_uring.c Signed-off-by: Li Lingfeng <lilingfeng3(a)huawei.com> Reviewed-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Wang Weiyang <wangweiyang2(a)huawei.com> Signed-off-by: Jialin Zhang <zhangjialin11(a)huawei.com> Signed-off-by: Zizhi Wo <wozizhi(a)huawei.com> --- io_uring/io_uring.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 8d1bc6cdfe71..d79e78971ec5 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -2440,6 +2440,9 @@ int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr) left--; break; } + /* to avoid doing too much in one submit round */ + if (left > IORING_MAX_ENTRIES / 2) + cond_resched(); } while (--left); if (unlikely(left)) { -- 2.39.2

2 3

[PATCH openEuler-1.0-LTS] x86/mce/amd: Publish the bank pointer only after setup has succeeded
by liwei 25 Nov '23

25 Nov '23

From: Borislav Petkov <bp(a)suse.de> mainline inclusion from mainline-v5.6-rc3 commit 6e5cf31fbe651bed7ba1df768f2e123531132417 category: bugfix bugzilla: 189381 CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… --------------------------- threshold_create_bank() creates a bank descriptor per MCA error thresholding counter which can be controlled over sysfs. It publishes the pointer to that bank in a per-CPU variable and then goes on to create additional thresholding blocks if the bank has such. However, that creation of additional blocks in allocate_threshold_blocks() can fail, leading to a use-after-free through the per-CPU pointer. Therefore, publish that pointer only after all blocks have been setup successfully. Fixes: 019f34fccfd5 ("x86, MCE, AMD: Move shared bank to node descriptor") Reported-by: Saar Amar <Saar.Amar(a)microsoft.com> Reported-by: Dan Carpenter <dan.carpenter(a)oracle.com> Signed-off-by: Borislav Petkov <bp(a)suse.de> Cc: <stable(a)vger.kernel.org> Link: http://lkml.kernel.org/r/20200128140846.phctkvx5btiexvbx@kili.mountain Signed-off-by: liwei <liwei728(a)huawei.com> --- arch/x86/kernel/cpu/mce/amd.c | 33 ++++++++++++++++----------------- 1 file changed, 16 insertions(+), 17 deletions(-) diff --git a/arch/x86/kernel/cpu/mce/amd.c b/arch/x86/kernel/cpu/mce/amd.c index 38b18fa055ef..af4a83695970 100644 --- a/arch/x86/kernel/cpu/mce/amd.c +++ b/arch/x86/kernel/cpu/mce/amd.c @@ -1200,8 +1200,9 @@ static const char *get_name(unsigned int bank, struct threshold_block *b) return buf_mcatype; } -static int allocate_threshold_blocks(unsigned int cpu, unsigned int bank, - unsigned int block, u32 address) +static int allocate_threshold_blocks(unsigned int cpu, struct threshold_bank *tb, + unsigned int bank, unsigned int block, + u32 address) { struct threshold_block *b = NULL; u32 low, high; @@ -1245,16 +1246,12 @@ static int allocate_threshold_blocks(unsigned int cpu, unsigned int bank, INIT_LIST_HEAD(&b->miscj); - if (per_cpu(threshold_banks, cpu)[bank]->blocks) { - list_add(&b->miscj, - &per_cpu(threshold_banks, cpu)[bank]->blocks->miscj); - } else { - per_cpu(threshold_banks, cpu)[bank]->blocks = b; - } + if (tb->blocks) + list_add(&b->miscj, &tb->blocks->miscj); + else + tb->blocks = b; - err = kobject_init_and_add(&b->kobj, &threshold_ktype, - per_cpu(threshold_banks, cpu)[bank]->kobj, - get_name(bank, b)); + err = kobject_init_and_add(&b->kobj, &threshold_ktype, tb->kobj, get_name(bank, b)); if (err) goto out_free; recurse: @@ -1262,7 +1259,7 @@ static int allocate_threshold_blocks(unsigned int cpu, unsigned int bank, if (!address) return 0; - err = allocate_threshold_blocks(cpu, bank, block, address); + err = allocate_threshold_blocks(cpu, tb, bank, block, address); if (err) goto out_free; @@ -1347,8 +1344,6 @@ static int threshold_create_bank(unsigned int cpu, unsigned int bank) goto out_free; } - per_cpu(threshold_banks, cpu)[bank] = b; - if (is_shared_bank(bank)) { refcount_set(&b->cpus, 1); @@ -1359,9 +1354,13 @@ static int threshold_create_bank(unsigned int cpu, unsigned int bank) } } - err = allocate_threshold_blocks(cpu, bank, 0, msr_ops.misc(bank)); - if (!err) - goto out; + err = allocate_threshold_blocks(cpu, b, bank, 0, msr_ops.misc(bank)); + if (err) + goto out_free; + + per_cpu(threshold_banks, cpu)[bank] = b; + + return 0; out_free: kfree(b); -- 2.25.1

2 1

[PATCH OLK-5.10 0/4] Add script to check & update openeuler_defconfig
by Jialin Zhang 25 Nov '23

25 Nov '23

Jialin Zhang (3): Revert "kconfig: Add script to update openeuler_defconfig" config: update openeuler_defconfig for x86 config: update openeuler_defconfig for arm64 Xie XiuQi (1): kconfig: Add script to check & update openeuler_defconfig arch/arm64/configs/openeuler_defconfig | 60 +++++++++++++++++--------- arch/x86/configs/openeuler_defconfig | 39 +++++++++++------ scripts/kconfig/Makefile | 14 ++++++ scripts/kconfig/makeconfig.sh | 5 +++ 4 files changed, 86 insertions(+), 32 deletions(-) -- 2.25.1

2 5

[PATCH openEuler-1.0-LTS 0/2] Backport crypto bugfix
by Yi Yang 25 Nov '23

25 Nov '23

crypto bugfix and fix kabi broken. Herbert Xu (1): crypto: api - Use work queue in crypto_destroy_instance Yi Yang (1): crypto: fix kabi broken in struct crypto_instance crypto/algapi.c | 29 +++++++++++++++++++++++++++-- include/crypto/algapi.h | 6 ++++++ 2 files changed, 33 insertions(+), 2 deletions(-) -- 2.25.1

2 3

[PATCH OLK-5.10 v7 00/20] introduce smart_grid zone
by Yipeng Zou 25 Nov '23

25 Nov '23

The patch sets include two parts: 1. patch 1~15,20: Rebase smart_grid from openeuler-1.0-LTS to OLK-5.10 2. patch 16~19: introduce smart_grid zone qos and cpufreq Since v6: 1. check is timer active in affinity timer Since v5: 1. Fix race condition in smart_grid_governor_enable. 2. Fix race condition in stop_auto_affinity. 3. Fix some str to int input in smart_grid_level_write. Since v4: 1. Place the highest level task in current domain level itself in sched_grid_prefer_cpus Since v3: 1. fix CI warning Since v2: 1. static alloc sg_zone cpumask. 2. fix some warning Hui Tang (13): sched: Introduce smart grid scheduling strategy for cfs sched: fix smart grid usage count sched: fix WARN found by deadlock detect sched: Fix possible deadlock in tg_set_dynamic_affinity_mode sched: Fix negative count for jump label sched: Fix timer storm for smart grid sched: fix dereference NULL pointers sched: Fix memory leak on error branch sched: clear credit count in error branch sched: Adjust few parameters range for smart grid sched: Delete redundant updates to p->prefer_cpus sched: Fix memory leak for smart grid sched: Fix null pointer derefrence for sd->span Wang ShaoBo (2): sched: smart grid: init sched_grid_qos structure on QOS purpose config: enable CONFIG_QOS_SCHED_SMART_GRID by default Yipeng Zou (5): sched: introduce smart grid qos zone smart_grid: introduce /proc/pid/smart_grid_level smart_grid: introduce smart_grid_strategy_ctrl sysctl smart_grid: cpufreq: introduce smart_grid cpufreq control sched: smart grid: check is active in affinity timer arch/arm64/configs/openeuler_defconfig | 1 + drivers/cpufreq/cpufreq.c | 232 ++++++++++++ fs/proc/array.c | 13 + fs/proc/base.c | 78 ++++ include/linux/cpufreq.h | 11 + include/linux/sched.h | 22 ++ include/linux/sched/grid_qos.h | 135 +++++++ include/linux/sched/sysctl.h | 5 + init/Kconfig | 13 + kernel/fork.c | 15 +- kernel/sched/Makefile | 1 + kernel/sched/core.c | 160 +++++++- kernel/sched/fair.c | 500 ++++++++++++++++++++++++- kernel/sched/grid/Makefile | 2 + kernel/sched/grid/internal.h | 6 + kernel/sched/grid/power.c | 27 ++ kernel/sched/grid/qos.c | 273 ++++++++++++++ kernel/sched/grid/stat.c | 47 +++ kernel/sched/sched.h | 48 +++ kernel/sysctl.c | 22 +- mm/mempolicy.c | 12 +- 21 files changed, 1605 insertions(+), 18 deletions(-) create mode 100644 include/linux/sched/grid_qos.h create mode 100644 kernel/sched/grid/Makefile create mode 100644 kernel/sched/grid/internal.h create mode 100644 kernel/sched/grid/power.c create mode 100644 kernel/sched/grid/qos.c create mode 100644 kernel/sched/grid/stat.c -- 2.34.1

1 20

[PATCH openEuler-1.0-LTS] umh: fix memory leak on execve failure
by Wenyu Huang 25 Nov '23

25 Nov '23

From: Vincent Minet <v.minet(a)criteo.com> mainline inclusion from mainline commit db803036ada7 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8J9MA CVE: NA ------------------------------------------------- If a UMH process created by fork_usermode_blob() fails to execute, a pair of struct file allocated by umh_pipe_setup() will leak. Under normal conditions, the caller (like bpfilter) needs to manage the lifetime of the UMH and its two pipes. But when fork_usermode_blob() fails, the caller doesn't really have a way to know what needs to be done. It seems better to do the cleanup ourselves in this case. Fixes: 449325b52b7a ("umh: introduce fork_usermode_blob() helper") Signed-off-by: Vincent Minet <v.minet(a)criteo.com> Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> Signed-off-by: Wenyu Huang <huangwenyu5(a)huawei.com> --- kernel/umh.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/kernel/umh.c b/kernel/umh.c index 53611efb10cb..715b4368f9ed 100644 --- a/kernel/umh.c +++ b/kernel/umh.c @@ -495,6 +495,12 @@ static void umh_clean_and_save_pid(struct subprocess_info *info) { struct umh_info *umh_info = info->data; + /* cleanup if umh_pipe_setup() was successful but exec failed */ + if (info->pid && info->retval) { + fput(umh_info->pipe_to_umh); + fput(umh_info->pipe_from_umh); + } + argv_free(info->argv); umh_info->pid = info->pid; } -- 2.34.1

2 1

[PATCH OLK-6.6 v3] net: sched: sch_qfq: Use non-work-conserving warning handler
by Zhengchao Shao 25 Nov '23

25 Nov '23

From: Liu Jian <liujian56(a)huawei.com> mainline inclusion from mainline-v6.7-rc1 commit 6d25d1dc76bf5943a5c1f4bb74d66d5eac58eb77 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8J4OL Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- A helper function for printing non-work-conserving alarms is added in commit b00355db3f88 ("pkt_sched: sch_hfsc: sch_htb: Add non-work-conserving warning handler."). In this commit, use qdisc_warn_nonwc() instead of WARN_ONCE() to handle the non-work-conserving warning in qfq Qdisc. Signed-off-by: Liu Jian <liujian56(a)huawei.com> Link: https://lore.kernel.org/r/20231023064729.370649-1-liujian56@huawei.com Signed-off-by: Paolo Abeni <pabeni(a)redhat.com> Signed-off-by: Zhengchao Shao <shaozhengchao(a)huawei.com> --- net/sched/sch_qfq.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/sched/sch_qfq.c b/net/sched/sch_qfq.c index 546c10adcacd..5598f8be18ae 100644 --- a/net/sched/sch_qfq.c +++ b/net/sched/sch_qfq.c @@ -1003,7 +1003,7 @@ static inline struct sk_buff *qfq_peek_skb(struct qfq_aggregate *agg, *cl = list_first_entry(&agg->active, struct qfq_class, alist); skb = (*cl)->qdisc->ops->peek((*cl)->qdisc); if (skb == NULL) - WARN_ONCE(1, "qfq_dequeue: non-workconserving leaf\n"); + qdisc_warn_nonwc("qfq_dequeue", (*cl)->qdisc); else *len = qdisc_pkt_len(skb); -- 2.34.1

2 1

[PATCH openEuler-22.03-LTS 00/17] Add error handle for add_disk
by Li Nan 25 Nov '23

25 Nov '23

Christoph Hellwig (5): block: fold register_disk into device_add_disk block: call blk_integrity_add earlier in device_add_disk block: add the events* attributes to disk_attrs block: fix error unwinding in device_add_disk block: clear ->slave_dir when dropping the main slave_dir reference Luis Chamberlain (4): block: return errors from blk_integrity_add block: return errors from disk_alloc_events block: add error handling for device_add_disk / add_disk block: fix device_add_disk() kobject_create_and_add() error handling Tetsuo Handa (1): block: check minor range in device_add_disk() Yu Kuai (1): block: fix memory leak for elevator on add_disk failure Zhong Jinghua (6): block: return errors from blk_register_region block: Fix the kabi change in device_add_disk block: Fix the kabi change on blk_register_region block: call blk_get_queue earlier in __device_add_disk block: Fix minor range check in device_add_disk() block: Set memalloc_noio to false in the error path block/blk.h | 5 +- include/linux/genhd.h | 13 +++ block/blk-integrity.c | 12 ++- block/genhd.c | 246 +++++++++++++++++++++++++----------------- 4 files changed, 173 insertions(+), 103 deletions(-) -- 2.39.2

2 19

[PATCH openEuler-1.0-LTS] umh: fix memory leak on execve failure
by Wenyu Huang 25 Nov '23

25 Nov '23

From: Vincent Minet <v.minet(a)criteo.com> mainline inclusion from mainline-v5.10 commit db803036ada7 ("umh: fix memory leak on execve failure") category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8J9MA CVE: NA ------------------------------------------------- If a UMH process created by fork_usermode_blob() fails to execute, a pair of struct file allocated by umh_pipe_setup() will leak. Under normal conditions, the caller (like bpfilter) needs to manage the lifetime of the UMH and its two pipes. But when fork_usermode_blob() fails, the caller doesn't really have a way to know what needs to be done. It seems better to do the cleanup ourselves in this case. Fixes: 449325b52b7a ("umh: introduce fork_usermode_blob() helper") Signed-off-by: Vincent Minet <v.minet(a)criteo.com> Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> Signed-off-by: Wenyu Huang <huangwenyu5(a)huawei.com> --- kernel/umh.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/kernel/umh.c b/kernel/umh.c index 53611efb10cb..715b4368f9ed 100644 --- a/kernel/umh.c +++ b/kernel/umh.c @@ -495,6 +495,12 @@ static void umh_clean_and_save_pid(struct subprocess_info *info) { struct umh_info *umh_info = info->data; + /* cleanup if umh_pipe_setup() was successful but exec failed */ + if (info->pid && info->retval) { + fput(umh_info->pipe_to_umh); + fput(umh_info->pipe_from_umh); + } + argv_free(info->argv); umh_info->pid = info->pid; } -- 2.34.1

2 1

[PATCH OLK-6.6 0/1] checkpatch: Update link tags to fix ci warning
by Zheng Zengkai 24 Nov '23

24 Nov '23

Update link tags in checkpatch.pl script to fix openEuler kernel ci warning: WARNING:COMMIT_LOG_USE_LINK: Unknown link reference 'bugzilla:', use 'Link:' or 'Closes:' instead. Wei Li (1): checkpatch: Update check of link tags scripts/checkpatch.pl | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) -- 2.20.1

2 2

[PATCH] crypto: hisilicon/qm - remove incorrect type cast
by Zhiqi Song 24 Nov '23

24 Nov '23

From: Weili Qian <qianweili(a)huawei.com> maillist inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8IVG6 CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… ---------------------------------------------------------------------- The 'offset' type is unsigned long in 'struct debugfs_reg32', so type of values casts to unsigned long long is incorrect, and the values do not require type cast, remove them. Signed-off-by: Weili Qian <qianweili(a)huawei.com> Signed-off-by: Herbert Xu <herbert(a)gondor.apana.org.au> Signed-off-by: JiangShui Yang <yangjiangshui(a)h-partners.com> --- drivers/crypto/hisilicon/debugfs.c | 50 +++++++++++------------ drivers/crypto/hisilicon/zip/zip_main.c | 54 ++++++++++++------------- 2 files changed, 52 insertions(+), 52 deletions(-) diff --git a/drivers/crypto/hisilicon/debugfs.c b/drivers/crypto/hisilicon/debugfs.c index 88885268851b..848195515de8 100644 --- a/drivers/crypto/hisilicon/debugfs.c +++ b/drivers/crypto/hisilicon/debugfs.c @@ -53,34 +53,34 @@ static struct qm_dfx_item qm_dfx_files[] = { #define CNT_CYC_REGS_NUM 10 static const struct debugfs_reg32 qm_dfx_regs[] = { /* XXX_CNT are reading clear register */ - {"QM_ECC_1BIT_CNT ", 0x104000ull}, - {"QM_ECC_MBIT_CNT ", 0x104008ull}, - {"QM_DFX_MB_CNT ", 0x104018ull}, - {"QM_DFX_DB_CNT ", 0x104028ull}, - {"QM_DFX_SQE_CNT ", 0x104038ull}, - {"QM_DFX_CQE_CNT ", 0x104048ull}, - {"QM_DFX_SEND_SQE_TO_ACC_CNT ", 0x104050ull}, - {"QM_DFX_WB_SQE_FROM_ACC_CNT ", 0x104058ull}, - {"QM_DFX_ACC_FINISH_CNT ", 0x104060ull}, - {"QM_DFX_CQE_ERR_CNT ", 0x1040b4ull}, - {"QM_DFX_FUNS_ACTIVE_ST ", 0x200ull}, - {"QM_ECC_1BIT_INF ", 0x104004ull}, - {"QM_ECC_MBIT_INF ", 0x10400cull}, - {"QM_DFX_ACC_RDY_VLD0 ", 0x1040a0ull}, - {"QM_DFX_ACC_RDY_VLD1 ", 0x1040a4ull}, - {"QM_DFX_AXI_RDY_VLD ", 0x1040a8ull}, - {"QM_DFX_FF_ST0 ", 0x1040c8ull}, - {"QM_DFX_FF_ST1 ", 0x1040ccull}, - {"QM_DFX_FF_ST2 ", 0x1040d0ull}, - {"QM_DFX_FF_ST3 ", 0x1040d4ull}, - {"QM_DFX_FF_ST4 ", 0x1040d8ull}, - {"QM_DFX_FF_ST5 ", 0x1040dcull}, - {"QM_DFX_FF_ST6 ", 0x1040e0ull}, - {"QM_IN_IDLE_ST ", 0x1040e4ull}, + {"QM_ECC_1BIT_CNT ", 0x104000}, + {"QM_ECC_MBIT_CNT ", 0x104008}, + {"QM_DFX_MB_CNT ", 0x104018}, + {"QM_DFX_DB_CNT ", 0x104028}, + {"QM_DFX_SQE_CNT ", 0x104038}, + {"QM_DFX_CQE_CNT ", 0x104048}, + {"QM_DFX_SEND_SQE_TO_ACC_CNT ", 0x104050}, + {"QM_DFX_WB_SQE_FROM_ACC_CNT ", 0x104058}, + {"QM_DFX_ACC_FINISH_CNT ", 0x104060}, + {"QM_DFX_CQE_ERR_CNT ", 0x1040b4}, + {"QM_DFX_FUNS_ACTIVE_ST ", 0x200}, + {"QM_ECC_1BIT_INF ", 0x104004}, + {"QM_ECC_MBIT_INF ", 0x10400c}, + {"QM_DFX_ACC_RDY_VLD0 ", 0x1040a0}, + {"QM_DFX_ACC_RDY_VLD1 ", 0x1040a4}, + {"QM_DFX_AXI_RDY_VLD ", 0x1040a8}, + {"QM_DFX_FF_ST0 ", 0x1040c8}, + {"QM_DFX_FF_ST1 ", 0x1040cc}, + {"QM_DFX_FF_ST2 ", 0x1040d0}, + {"QM_DFX_FF_ST3 ", 0x1040d4}, + {"QM_DFX_FF_ST4 ", 0x1040d8}, + {"QM_DFX_FF_ST5 ", 0x1040dc}, + {"QM_DFX_FF_ST6 ", 0x1040e0}, + {"QM_IN_IDLE_ST ", 0x1040e4}, }; static const struct debugfs_reg32 qm_vf_dfx_regs[] = { - {"QM_DFX_FUNS_ACTIVE_ST ", 0x200ull}, + {"QM_DFX_FUNS_ACTIVE_ST ", 0x200}, }; /* define the QM's dfx regs region and region length */ diff --git a/drivers/crypto/hisilicon/zip/zip_main.c b/drivers/crypto/hisilicon/zip/zip_main.c index e8dbf7759c80..944afeb9da73 100644 --- a/drivers/crypto/hisilicon/zip/zip_main.c +++ b/drivers/crypto/hisilicon/zip/zip_main.c @@ -285,28 +285,28 @@ static const u64 core_offsets[] = { }; static const struct debugfs_reg32 hzip_dfx_regs[] = { - {"HZIP_GET_BD_NUM ", 0x00ull}, - {"HZIP_GET_RIGHT_BD ", 0x04ull}, - {"HZIP_GET_ERROR_BD ", 0x08ull}, - {"HZIP_DONE_BD_NUM ", 0x0cull}, - {"HZIP_WORK_CYCLE ", 0x10ull}, - {"HZIP_IDLE_CYCLE ", 0x18ull}, - {"HZIP_MAX_DELAY ", 0x20ull}, - {"HZIP_MIN_DELAY ", 0x24ull}, - {"HZIP_AVG_DELAY ", 0x28ull}, - {"HZIP_MEM_VISIBLE_DATA ", 0x30ull}, - {"HZIP_MEM_VISIBLE_ADDR ", 0x34ull}, - {"HZIP_CONSUMED_BYTE ", 0x38ull}, - {"HZIP_PRODUCED_BYTE ", 0x40ull}, - {"HZIP_COMP_INF ", 0x70ull}, - {"HZIP_PRE_OUT ", 0x78ull}, - {"HZIP_BD_RD ", 0x7cull}, - {"HZIP_BD_WR ", 0x80ull}, - {"HZIP_GET_BD_AXI_ERR_NUM ", 0x84ull}, - {"HZIP_GET_BD_PARSE_ERR_NUM ", 0x88ull}, - {"HZIP_ADD_BD_AXI_ERR_NUM ", 0x8cull}, - {"HZIP_DECOMP_STF_RELOAD_CURR_ST ", 0x94ull}, - {"HZIP_DECOMP_LZ77_CURR_ST ", 0x9cull}, + {"HZIP_GET_BD_NUM ", 0x00}, + {"HZIP_GET_RIGHT_BD ", 0x04}, + {"HZIP_GET_ERROR_BD ", 0x08}, + {"HZIP_DONE_BD_NUM ", 0x0c}, + {"HZIP_WORK_CYCLE ", 0x10}, + {"HZIP_IDLE_CYCLE ", 0x18}, + {"HZIP_MAX_DELAY ", 0x20}, + {"HZIP_MIN_DELAY ", 0x24}, + {"HZIP_AVG_DELAY ", 0x28}, + {"HZIP_MEM_VISIBLE_DATA ", 0x30}, + {"HZIP_MEM_VISIBLE_ADDR ", 0x34}, + {"HZIP_CONSUMED_BYTE ", 0x38}, + {"HZIP_PRODUCED_BYTE ", 0x40}, + {"HZIP_COMP_INF ", 0x70}, + {"HZIP_PRE_OUT ", 0x78}, + {"HZIP_BD_RD ", 0x7c}, + {"HZIP_BD_WR ", 0x80}, + {"HZIP_GET_BD_AXI_ERR_NUM ", 0x84}, + {"HZIP_GET_BD_PARSE_ERR_NUM ", 0x88}, + {"HZIP_ADD_BD_AXI_ERR_NUM ", 0x8c}, + {"HZIP_DECOMP_STF_RELOAD_CURR_ST ", 0x94}, + {"HZIP_DECOMP_LZ77_CURR_ST ", 0x9c}, }; static const struct debugfs_reg32 hzip_com_dfx_regs[] = { @@ -318,11 +318,11 @@ static const struct debugfs_reg32 hzip_com_dfx_regs[] = { }; static const struct debugfs_reg32 hzip_dump_dfx_regs[] = { - {"HZIP_GET_BD_NUM ", 0x00ull}, - {"HZIP_GET_RIGHT_BD ", 0x04ull}, - {"HZIP_GET_ERROR_BD ", 0x08ull}, - {"HZIP_DONE_BD_NUM ", 0x0cull}, - {"HZIP_MAX_DELAY ", 0x20ull}, + {"HZIP_GET_BD_NUM ", 0x00}, + {"HZIP_GET_RIGHT_BD ", 0x04}, + {"HZIP_GET_ERROR_BD ", 0x08}, + {"HZIP_DONE_BD_NUM ", 0x0c}, + {"HZIP_MAX_DELAY ", 0x20}, }; /* define the ZIP's dfx regs region and region length */ -- 2.30.0

1 0

[PATCH] crypto: hisilicon/qm - print device abnormal information
by Zhiqi Song 24 Nov '23

24 Nov '23

From: Weili Qian <qianweili(a)huawei.com> maillist inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8IVG6 CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… ---------------------------------------------------------------------- When device is abnormal and reports abnormal interrupt event to driver, the driver can print device information for error analysis. This patch adds some device error-related information output after the device reports an abnormal interrupt. Signed-off-by: Weili Qian <qianweili(a)huawei.com> Signed-off-by: Herbert Xu <herbert(a)gondor.apana.org.au> Signed-off-by: JiangShui Yang <yangjiangshui(a)h-partners.com> --- drivers/crypto/hisilicon/qm.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c index 8f99683bcdb5..b7eab69318f5 100644 --- a/drivers/crypto/hisilicon/qm.c +++ b/drivers/crypto/hisilicon/qm.c @@ -131,16 +131,21 @@ #define QM_FIFO_OVERFLOW_TYPE 0xc0 #define QM_FIFO_OVERFLOW_TYPE_SHIFT 6 #define QM_FIFO_OVERFLOW_VF 0x3f +#define QM_FIFO_OVERFLOW_QP_SHIFT 16 #define QM_ABNORMAL_INF01 0x100014 #define QM_DB_TIMEOUT_TYPE 0xc0 #define QM_DB_TIMEOUT_TYPE_SHIFT 6 #define QM_DB_TIMEOUT_VF 0x3f +#define QM_DB_TIMEOUT_QP_SHIFT 16 +#define QM_ABNORMAL_INF02 0x100018 +#define QM_AXI_POISON_ERR BIT(22) #define QM_RAS_CE_ENABLE 0x1000ec #define QM_RAS_FE_ENABLE 0x1000f0 #define QM_RAS_NFE_ENABLE 0x1000f4 #define QM_RAS_CE_THRESHOLD 0x1000f8 #define QM_RAS_CE_TIMES_PER_IRQ 1 #define QM_OOO_SHUTDOWN_SEL 0x1040f8 +#define QM_AXI_RRESP_ERR BIT(0) #define QM_ECC_MBIT BIT(2) #define QM_DB_TIMEOUT BIT(10) #define QM_OF_FIFO_OF BIT(11) @@ -1538,7 +1543,7 @@ static void qm_log_hw_error(struct hisi_qm *qm, u32 error_status) { const struct hisi_qm_hw_error *err; struct device *dev = &qm->pdev->dev; - u32 reg_val, type, vf_num; + u32 reg_val, type, vf_num, qp_id; int i; for (i = 0; i < ARRAY_SIZE(qm_hw_error); i++) { @@ -1554,19 +1559,24 @@ static void qm_log_hw_error(struct hisi_qm *qm, u32 error_status) type = (reg_val & QM_DB_TIMEOUT_TYPE) >> QM_DB_TIMEOUT_TYPE_SHIFT; vf_num = reg_val & QM_DB_TIMEOUT_VF; - dev_err(dev, "qm %s doorbell timeout in function %u\n", - qm_db_timeout[type], vf_num); + qp_id = reg_val >> QM_DB_TIMEOUT_QP_SHIFT; + dev_err(dev, "qm %s doorbell timeout in function %u qp %u\n", + qm_db_timeout[type], vf_num, qp_id); } else if (err->int_msk & QM_OF_FIFO_OF) { reg_val = readl(qm->io_base + QM_ABNORMAL_INF00); type = (reg_val & QM_FIFO_OVERFLOW_TYPE) >> QM_FIFO_OVERFLOW_TYPE_SHIFT; vf_num = reg_val & QM_FIFO_OVERFLOW_VF; - + qp_id = reg_val >> QM_FIFO_OVERFLOW_QP_SHIFT; if (type < ARRAY_SIZE(qm_fifo_overflow)) - dev_err(dev, "qm %s fifo overflow in function %u\n", - qm_fifo_overflow[type], vf_num); + dev_err(dev, "qm %s fifo overflow in function %u qp %u\n", + qm_fifo_overflow[type], vf_num, qp_id); else dev_err(dev, "unknown error type\n"); + } else if (err->int_msk & QM_AXI_RRESP_ERR) { + reg_val = readl(qm->io_base + QM_ABNORMAL_INF02); + if (reg_val & QM_AXI_POISON_ERR) + dev_err(dev, "qm axi poison error happened\n"); } } } -- 2.30.0

1 0

[PATCH] crypto: hisilicon/qm - check function qp num before alg register
by Zhiqi Song 24 Nov '23

24 Nov '23

From: Weili Qian <qianweili(a)huawei.com> mainline inclusion from mainline-v6.7-rc1 commit b42ab1c61a77832040ad42ebf9adf237360e49f7 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8IVG6 CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… ---------------------------------------------------------------------- When the Kunpeng accelerator executes tasks such as encryption and decryption have minimum requirements on the number of device queues. If the number of queues does not meet the requirement, the process initialization will fail. Therefore, the driver checks the number of queues on the device before registering the algorithm. If the number does not meet the requirements, the driver does not register the algorithm to crypto subsystem, the device is still added to the qm_list. Signed-off-by: Weili Qian <qianweili(a)huawei.com> Signed-off-by: Herbert Xu <herbert(a)gondor.apana.org.au> Signed-off-by: JiangShui Yang <yangjiangshui(a)h-partners.com> --- drivers/crypto/hisilicon/hpre/hpre_crypto.c | 25 ++++++++++- drivers/crypto/hisilicon/hpre/hpre_main.c | 14 +++--- drivers/crypto/hisilicon/qm.c | 47 +++++++-------------- drivers/crypto/hisilicon/sec2/sec_crypto.c | 31 ++++++++++++-- drivers/crypto/hisilicon/sec2/sec_main.c | 24 +++++------ drivers/crypto/hisilicon/zip/zip_crypto.c | 23 +++++++++- drivers/crypto/hisilicon/zip/zip_main.c | 14 +++--- include/linux/hisi_acc_qm.h | 18 +++++++- 8 files changed, 133 insertions(+), 63 deletions(-) diff --git a/drivers/crypto/hisilicon/hpre/hpre_crypto.c b/drivers/crypto/hisilicon/hpre/hpre_crypto.c index b0a7eb3b6362..1f64e63b75cb 100644 --- a/drivers/crypto/hisilicon/hpre/hpre_crypto.c +++ b/drivers/crypto/hisilicon/hpre/hpre_crypto.c @@ -57,6 +57,9 @@ struct hpre_ctx; #define HPRE_DRV_ECDH_MASK_CAP BIT(2) #define HPRE_DRV_X25519_MASK_CAP BIT(5) +static DEFINE_MUTEX(hpre_algs_lock); +static unsigned int hpre_available_devs; + typedef void (*hpre_cb)(struct hpre_ctx *ctx, void *sqe); struct hpre_rsa_ctx { @@ -2192,11 +2195,17 @@ static void hpre_unregister_x25519(struct hisi_qm *qm) int hpre_algs_register(struct hisi_qm *qm) { - int ret; + int ret = 0; + + mutex_lock(&hpre_algs_lock); + if (hpre_available_devs) { + hpre_available_devs++; + goto unlock; + } ret = hpre_register_rsa(qm); if (ret) - return ret; + goto unlock; ret = hpre_register_dh(qm); if (ret) @@ -2210,6 +2219,9 @@ int hpre_algs_register(struct hisi_qm *qm) if (ret) goto unreg_ecdh; + hpre_available_devs++; + mutex_unlock(&hpre_algs_lock); + return ret; unreg_ecdh: @@ -2218,13 +2230,22 @@ int hpre_algs_register(struct hisi_qm *qm) hpre_unregister_dh(qm); unreg_rsa: hpre_unregister_rsa(qm); +unlock: + mutex_unlock(&hpre_algs_lock); return ret; } void hpre_algs_unregister(struct hisi_qm *qm) { + mutex_lock(&hpre_algs_lock); + if (--hpre_available_devs) + goto unlock; + hpre_unregister_x25519(qm); hpre_unregister_ecdh(qm); hpre_unregister_dh(qm); hpre_unregister_rsa(qm); + +unlock: + mutex_unlock(&hpre_algs_lock); } diff --git a/drivers/crypto/hisilicon/hpre/hpre_main.c b/drivers/crypto/hisilicon/hpre/hpre_main.c index 005e8c46ce35..580159cb6582 100644 --- a/drivers/crypto/hisilicon/hpre/hpre_main.c +++ b/drivers/crypto/hisilicon/hpre/hpre_main.c @@ -107,6 +107,7 @@ #define HPRE_VIA_MSI_DSM 1 #define HPRE_SQE_MASK_OFFSET 8 #define HPRE_SQE_MASK_LEN 24 +#define HPRE_CTX_Q_NUM_DEF 1 #define HPRE_DFX_BASE 0x301000 #define HPRE_DFX_COMMON1 0x301400 @@ -1382,10 +1383,11 @@ static int hpre_probe(struct pci_dev *pdev, const struct pci_device_id *id) if (ret) dev_warn(&pdev->dev, "init debugfs fail!\n"); - ret = hisi_qm_alg_register(qm, &hpre_devices); + hisi_qm_add_list(qm, &hpre_devices); + ret = hisi_qm_alg_register(qm, &hpre_devices, HPRE_CTX_Q_NUM_DEF); if (ret < 0) { pci_err(pdev, "fail to register algs to crypto!\n"); - goto err_with_qm_start; + goto err_qm_del_list; } ret = qm_register_uacce(qm); @@ -1405,9 +1407,10 @@ static int hpre_probe(struct pci_dev *pdev, const struct pci_device_id *id) return 0; err_with_alg_register: - hisi_qm_alg_unregister(qm, &hpre_devices); + hisi_qm_alg_unregister(qm, &hpre_devices, HPRE_CTX_Q_NUM_DEF); -err_with_qm_start: +err_qm_del_list: + hisi_qm_del_list(qm, &hpre_devices); hpre_debugfs_exit(qm); hisi_qm_stop(qm, QM_NORMAL); @@ -1427,7 +1430,8 @@ static void hpre_remove(struct pci_dev *pdev) hisi_qm_pm_uninit(qm); hisi_qm_wait_task_finish(qm, &hpre_devices); - hisi_qm_alg_unregister(qm, &hpre_devices); + hisi_qm_alg_unregister(qm, &hpre_devices, HPRE_CTX_Q_NUM_DEF); + hisi_qm_del_list(qm, &hpre_devices); if (qm->fun_type == QM_HW_PF && qm->vfs_num) hisi_qm_sriov_disable(pdev, true); diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c index d67d2d83d590..2c45d5ecce69 100644 --- a/drivers/crypto/hisilicon/qm.c +++ b/drivers/crypto/hisilicon/qm.c @@ -5027,63 +5027,48 @@ static void qm_cmd_process(struct work_struct *cmd_process) } /** - * hisi_qm_alg_register() - Register alg to crypto and add qm to qm_list. + * hisi_qm_alg_register() - Register alg to crypto. * @qm: The qm needs add. * @qm_list: The qm list. + * @guard: Guard of qp_num. * - * This function adds qm to qm list, and will register algorithm to - * crypto when the qm list is empty. + * Register algorithm to crypto when the function is satisfy guard. */ -int hisi_qm_alg_register(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +int hisi_qm_alg_register(struct hisi_qm *qm, struct hisi_qm_list *qm_list, int guard) { struct device *dev = &qm->pdev->dev; - int flag = 0; - int ret = 0; - - mutex_lock(&qm_list->lock); - if (list_empty(&qm_list->list)) - flag = 1; - list_add_tail(&qm->list, &qm_list->list); - mutex_unlock(&qm_list->lock); if (qm->ver <= QM_HW_V2 && qm->use_sva) { dev_info(dev, "HW V2 not both use uacce sva mode and hardware crypto algs.\n"); return 0; } - if (flag) { - ret = qm_list->register_to_crypto(qm); - if (ret) { - mutex_lock(&qm_list->lock); - list_del(&qm->list); - mutex_unlock(&qm_list->lock); - } + if (qm->qp_num < guard) { + dev_info(dev, "qp_num is less than task need.\n"); + return 0; } - return ret; + return qm_list->register_to_crypto(qm); } EXPORT_SYMBOL_GPL(hisi_qm_alg_register); /** - * hisi_qm_alg_unregister() - Unregister alg from crypto and delete qm from - * qm list. + * hisi_qm_alg_unregister() - Unregister alg from crypto. * @qm: The qm needs delete. * @qm_list: The qm list. + * @guard: Guard of qp_num. * - * This function deletes qm from qm list, and will unregister algorithm - * from crypto when the qm list is empty. + * Unregister algorithm from crypto when the last function is satisfy guard. */ -void hisi_qm_alg_unregister(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +void hisi_qm_alg_unregister(struct hisi_qm *qm, struct hisi_qm_list *qm_list, int guard) { - mutex_lock(&qm_list->lock); - list_del(&qm->list); - mutex_unlock(&qm_list->lock); - if (qm->ver <= QM_HW_V2 && qm->use_sva) return; - if (list_empty(&qm_list->list)) - qm_list->unregister_from_crypto(qm); + if (qm->qp_num < guard) + return; + + qm_list->unregister_from_crypto(qm); } EXPORT_SYMBOL_GPL(hisi_qm_alg_unregister); diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c index 27db9fa65ddf..4d30d49c605e 100644 --- a/drivers/crypto/hisilicon/sec2/sec_crypto.c +++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c @@ -104,6 +104,9 @@ #define IV_CTR_INIT 0x1 #define IV_BYTE_OFFSET 0x8 +static DEFINE_MUTEX(sec_algs_lock); +static unsigned int sec_available_devs; + struct sec_skcipher { u64 alg_msk; struct skcipher_alg alg; @@ -2536,16 +2539,31 @@ static int sec_register_aead(u64 alg_mask) int sec_register_to_crypto(struct hisi_qm *qm) { u64 alg_mask = sec_get_alg_bitmap(qm, SEC_DRV_ALG_BITMAP_HIGH, SEC_DRV_ALG_BITMAP_LOW); - int ret; + int ret = 0; + + mutex_lock(&sec_algs_lock); + if (sec_available_devs) { + sec_available_devs++; + goto unlock; + } ret = sec_register_skcipher(alg_mask); if (ret) - return ret; + goto unlock; ret = sec_register_aead(alg_mask); if (ret) - sec_unregister_skcipher(alg_mask, ARRAY_SIZE(sec_skciphers)); + goto unreg_skcipher; + sec_available_devs++; + mutex_unlock(&sec_algs_lock); + + return 0; + +unreg_skcipher: + sec_unregister_skcipher(alg_mask, ARRAY_SIZE(sec_skciphers)); +unlock: + mutex_unlock(&sec_algs_lock); return ret; } @@ -2553,6 +2571,13 @@ void sec_unregister_from_crypto(struct hisi_qm *qm) { u64 alg_mask = sec_get_alg_bitmap(qm, SEC_DRV_ALG_BITMAP_HIGH, SEC_DRV_ALG_BITMAP_LOW); + mutex_lock(&sec_algs_lock); + if (--sec_available_devs) + goto unlock; + sec_unregister_aead(alg_mask, ARRAY_SIZE(sec_aeads)); sec_unregister_skcipher(alg_mask, ARRAY_SIZE(sec_skciphers)); + +unlock: + mutex_unlock(&sec_algs_lock); } diff --git a/drivers/crypto/hisilicon/sec2/sec_main.c b/drivers/crypto/hisilicon/sec2/sec_main.c index 09831a8c6fad..29461754490e 100644 --- a/drivers/crypto/hisilicon/sec2/sec_main.c +++ b/drivers/crypto/hisilicon/sec2/sec_main.c @@ -1205,15 +1205,11 @@ static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id) if (ret) pci_warn(pdev, "Failed to init debugfs!\n"); - if (qm->qp_num >= ctx_q_num) { - ret = hisi_qm_alg_register(qm, &sec_devices); - if (ret < 0) { - pr_err("Failed to register driver to crypto.\n"); - goto err_qm_stop; - } - } else { - pci_warn(qm->pdev, - "Failed to use kernel mode, qp not enough!\n"); + hisi_qm_add_list(qm, &sec_devices); + ret = hisi_qm_alg_register(qm, &sec_devices, ctx_q_num); + if (ret < 0) { + pr_err("Failed to register driver to crypto.\n"); + goto err_qm_del_list; } if (qm->uacce) { @@ -1235,9 +1231,9 @@ static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id) return 0; err_alg_unregister: - if (qm->qp_num >= ctx_q_num) - hisi_qm_alg_unregister(qm, &sec_devices); -err_qm_stop: + hisi_qm_alg_unregister(qm, &sec_devices, ctx_q_num); +err_qm_del_list: + hisi_qm_del_list(qm, &sec_devices); sec_debugfs_exit(qm); hisi_qm_stop(qm, QM_NORMAL); err_probe_uninit: @@ -1254,8 +1250,8 @@ static void sec_remove(struct pci_dev *pdev) hisi_qm_pm_uninit(qm); hisi_qm_wait_task_finish(qm, &sec_devices); - if (qm->qp_num >= ctx_q_num) - hisi_qm_alg_unregister(qm, &sec_devices); + hisi_qm_alg_unregister(qm, &sec_devices, ctx_q_num); + hisi_qm_del_list(qm, &sec_devices); if (qm->fun_type == QM_HW_PF && qm->vfs_num) hisi_qm_sriov_disable(pdev, true); diff --git a/drivers/crypto/hisilicon/zip/zip_crypto.c b/drivers/crypto/hisilicon/zip/zip_crypto.c index 09f60f786779..304411ac5a36 100644 --- a/drivers/crypto/hisilicon/zip/zip_crypto.c +++ b/drivers/crypto/hisilicon/zip/zip_crypto.c @@ -44,6 +44,9 @@ #define HZIP_ALG_GZIP GENMASK(3, 2) #define HZIP_ALG_DEFLATE GENMASK(5, 4) +static DEFINE_MUTEX(zip_algs_lock); +static unsigned int zip_available_devs; + static const u8 zlib_head[HZIP_ZLIB_HEAD_SIZE] = {0x78, 0x9c}; static const u8 gzip_head[HZIP_GZIP_HEAD_SIZE] = { 0x1f, 0x8b, 0x08, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x03 @@ -864,9 +867,15 @@ int hisi_zip_register_to_crypto(struct hisi_qm *qm) { int ret = 0; + mutex_lock(&zip_algs_lock); + if (zip_available_devs) { + zip_available_devs++; + goto err_unlock; + } + ret = hisi_zip_register_deflate(qm); if (ret) - return ret; + goto err_unlock; ret = hisi_zip_register_zlib(qm); if (ret) @@ -876,18 +885,30 @@ int hisi_zip_register_to_crypto(struct hisi_qm *qm) if (ret) goto err_unreg_zlib; + zip_available_devs++; + mutex_unlock(&zip_algs_lock); + return 0; err_unreg_zlib: hisi_zip_unregister_zlib(qm); err_unreg_deflate: hisi_zip_unregister_deflate(qm); +err_unlock: + mutex_unlock(&zip_algs_lock); return ret; } void hisi_zip_unregister_from_crypto(struct hisi_qm *qm) { + mutex_lock(&zip_algs_lock); + if (--zip_available_devs) + goto unlock; + hisi_zip_unregister_deflate(qm); hisi_zip_unregister_zlib(qm); hisi_zip_unregister_gzip(qm); + +unlock: + mutex_unlock(&zip_algs_lock); } diff --git a/drivers/crypto/hisilicon/zip/zip_main.c b/drivers/crypto/hisilicon/zip/zip_main.c index 2d56a546c76f..7bb512e3cb32 100644 --- a/drivers/crypto/hisilicon/zip/zip_main.c +++ b/drivers/crypto/hisilicon/zip/zip_main.c @@ -67,6 +67,7 @@ #define HZIP_SQE_SIZE 128 #define HZIP_PF_DEF_Q_NUM 64 #define HZIP_PF_DEF_Q_BASE 0 +#define HZIP_CTX_Q_NUM_DEF 2 #define HZIP_SOFT_CTRL_CNT_CLR_CE 0x301000 #define HZIP_SOFT_CTRL_CNT_CLR_CE_BIT BIT(0) @@ -1215,10 +1216,11 @@ static int hisi_zip_probe(struct pci_dev *pdev, const struct pci_device_id *id) if (ret) pci_err(pdev, "failed to init debugfs (%d)!\n", ret); - ret = hisi_qm_alg_register(qm, &zip_devices); + hisi_qm_add_list(qm, &zip_devices); + ret = hisi_qm_alg_register(qm, &zip_devices, HZIP_CTX_Q_NUM_DEF); if (ret < 0) { pci_err(pdev, "failed to register driver to crypto!\n"); - goto err_qm_stop; + goto err_qm_del_list; } if (qm->uacce) { @@ -1240,9 +1242,10 @@ static int hisi_zip_probe(struct pci_dev *pdev, const struct pci_device_id *id) return 0; err_qm_alg_unregister: - hisi_qm_alg_unregister(qm, &zip_devices); + hisi_qm_alg_unregister(qm, &zip_devices, HZIP_CTX_Q_NUM_DEF); -err_qm_stop: +err_qm_del_list: + hisi_qm_del_list(qm, &zip_devices); hisi_zip_debugfs_exit(qm); hisi_qm_stop(qm, QM_NORMAL); @@ -1262,7 +1265,8 @@ static void hisi_zip_remove(struct pci_dev *pdev) hisi_qm_pm_uninit(qm); hisi_qm_wait_task_finish(qm, &zip_devices); - hisi_qm_alg_unregister(qm, &zip_devices); + hisi_qm_alg_unregister(qm, &zip_devices, HZIP_CTX_Q_NUM_DEF); + hisi_qm_del_list(qm, &zip_devices); if (qm->fun_type == QM_HW_PF && qm->vfs_num) hisi_qm_sriov_disable(pdev, true); diff --git a/include/linux/hisi_acc_qm.h b/include/linux/hisi_acc_qm.h index 03860e33f8eb..b0147be3a606 100644 --- a/include/linux/hisi_acc_qm.h +++ b/include/linux/hisi_acc_qm.h @@ -519,6 +519,20 @@ static inline void hisi_qm_init_list(struct hisi_qm_list *qm_list) mutex_init(&qm_list->lock); } +static inline void hisi_qm_add_list(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +{ + mutex_lock(&qm_list->lock); + list_add_tail(&qm->list, &qm_list->list); + mutex_unlock(&qm_list->lock); +} + +static inline void hisi_qm_del_list(struct hisi_qm *qm, struct hisi_qm_list *qm_list) +{ + mutex_lock(&qm_list->lock); + list_del(&qm->list); + mutex_unlock(&qm_list->lock); +} + int qm_register_uacce(struct hisi_qm *qm); int hisi_qm_init(struct hisi_qm *qm); void hisi_qm_uninit(struct hisi_qm *qm); @@ -561,8 +575,8 @@ int hisi_qm_alloc_qps_node(struct hisi_qm_list *qm_list, int qp_num, void hisi_qm_free_qps(struct hisi_qp **qps, int qp_num); void hisi_qm_dev_shutdown(struct pci_dev *pdev); void hisi_qm_wait_task_finish(struct hisi_qm *qm, struct hisi_qm_list *qm_list); -int hisi_qm_alg_register(struct hisi_qm *qm, struct hisi_qm_list *qm_list); -void hisi_qm_alg_unregister(struct hisi_qm *qm, struct hisi_qm_list *qm_list); +int hisi_qm_alg_register(struct hisi_qm *qm, struct hisi_qm_list *qm_list, int guard); +void hisi_qm_alg_unregister(struct hisi_qm *qm, struct hisi_qm_list *qm_list, int guard); int hisi_qm_resume(struct device *dev); int hisi_qm_suspend(struct device *dev); void hisi_qm_pm_uninit(struct hisi_qm *qm); -- 2.30.0

1 0

[PATCH 0/5] crypto: hisilicon - fix the process to obtain capability register value
by Zhiqi Song 24 Nov '23

24 Nov '23

This seires patch modify the process to obtain the value of capablity registers. Pre-store the valid values of them. Wenkai Lin (1): crypto: hisilicon/qm - add a function to set qm algs Zhiqi Song (4): crypto: hisilicon/qm - save capability registers in qm init process crypto: hisilicon/hpre - save capability registers in probe process crypto: hisilicon/sec2 - save capability registers in probe process crypto: hisilicon/zip - save capability registers in probe process drivers/crypto/hisilicon/hpre/hpre_main.c | 122 +++++++++++---------- drivers/crypto/hisilicon/qm.c | 98 +++++++++++++++-- drivers/crypto/hisilicon/sec2/sec.h | 7 ++ drivers/crypto/hisilicon/sec2/sec_crypto.c | 8 +- drivers/crypto/hisilicon/sec2/sec_main.c | 78 ++++++------- drivers/crypto/hisilicon/zip/zip_main.c | 120 +++++++++++--------- include/linux/hisi_acc_qm.h | 20 +++- 7 files changed, 295 insertions(+), 158 deletions(-) -- 2.30.0

1 5

【邀请函】操作系统大会&openEuler Summit 2023
by openEuler 24 Nov '23

24 Nov '23

1 0

[PATCH openEuler-1.0-LTS] umh: fix memory leak on execve failure
by Wenyu Huang 24 Nov '23

24 Nov '23

From: Vincent Minet <v.minet(a)criteo.com> mainline inclusion from mainline-5.10 commit db803036ada7 ("umh: fix memory leak on execve failure") category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8J9MA CVE: NA ------------------------------------------------- If a UMH process created by fork_usermode_blob() fails to execute, a pair of struct file allocated by umh_pipe_setup() will leak. Under normal conditions, the caller (like bpfilter) needs to manage the lifetime of the UMH and its two pipes. But when fork_usermode_blob() fails, the caller doesn't really have a way to know what needs to be done. It seems better to do the cleanup ourselves in this case. Fixes: 449325b52b7a ("umh: introduce fork_usermode_blob() helper") Signed-off-by: Vincent Minet <v.minet(a)criteo.com> Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> Signed-off-by: Wenyu Huang <huangwenyu5(a)huawei.com> --- kernel/umh.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/kernel/umh.c b/kernel/umh.c index 53611efb10cb..715b4368f9ed 100644 --- a/kernel/umh.c +++ b/kernel/umh.c @@ -495,6 +495,12 @@ static void umh_clean_and_save_pid(struct subprocess_info *info) { struct umh_info *umh_info = info->data; + /* cleanup if umh_pipe_setup() was successful but exec failed */ + if (info->pid && info->retval) { + fput(umh_info->pipe_to_umh); + fput(umh_info->pipe_from_umh); + } + argv_free(info->argv); umh_info->pid = info->pid; } -- 2.34.1

2 1

[PATCH OLK-6.6 v2] net: sched: sch_qfq: Use non-work-conserving warning handler
by Zhengchao Shao 24 Nov '23

24 Nov '23

From: Liu Jian <liujian56(a)huawei.com> mainline inclusion from mainline-v6.7-rc1 commit 6d25d1dc76bf5943a5c1f4bb74d66d5eac58eb77 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8J4OL Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- A helper function for printing non-work-conserving alarms is added in commit b00355db3f88 ("pkt_sched: sch_hfsc: sch_htb: Add non-work-conserving warning handler."). In this commit, use qdisc_warn_nonwc() instead of WARN_ONCE() to handle the non-work-conserving warning in qfq Qdisc. Signed-off-by: Liu Jian <liujian56(a)huawei.com> Link: https://lore.kernel.org/r/20231023064729.370649-1-liujian56@huawei.com Signed-off-by: Paolo Abeni <pabeni(a)redhat.com> Signed-off-by: Zhengchao Shao <shaozhengchao(a)huawei.com> --- net/sched/sch_qfq.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/sched/sch_qfq.c b/net/sched/sch_qfq.c index 546c10adcacd..5598f8be18ae 100644 --- a/net/sched/sch_qfq.c +++ b/net/sched/sch_qfq.c @@ -1003,7 +1003,7 @@ static inline struct sk_buff *qfq_peek_skb(struct qfq_aggregate *agg, *cl = list_first_entry(&agg->active, struct qfq_class, alist); skb = (*cl)->qdisc->ops->peek((*cl)->qdisc); if (skb == NULL) - WARN_ONCE(1, "qfq_dequeue: non-workconserving leaf\n"); + qdisc_warn_nonwc("qfq_dequeue", (*cl)->qdisc); else *len = qdisc_pkt_len(skb); -- 2.34.1

2 1

[PATCH openEuler-1.0-LTS 0/2] Backport crypto bugfix
by Yi Yang 24 Nov '23

24 Nov '23

crypto bugfix and fix kabi broken. Herbert Xu (1): crypto: api - Use work queue in crypto_destroy_instance Yi Yang (1): crypto: fix kabi broken in struct crypto_instance crypto/algapi.c | 29 +++++++++++++++++++++++++++-- include/crypto/algapi.h | 5 +++++ 2 files changed, 32 insertions(+), 2 deletions(-) -- 2.25.1

2 3

[PATCH OLK-5.10 0/2] Backport crypto bugfix
by Yi Yang 24 Nov '23

24 Nov '23

crypto bugfix and fix kabi broken. Herbert Xu (1): crypto: api - Use work queue in crypto_destroy_instance Yi Yang (1): crypto: fix kabi broken in struct crypto_instance crypto/algapi.c | 29 +++++++++++++++++++++++++++-- include/crypto/algapi.h | 6 ++++++ 2 files changed, 33 insertions(+), 2 deletions(-) -- 2.25.1

2 3

[PATCH OLK-5.10] time/sched_clock: Remove clock_read_data's read mostly members
by Zeng Heng 24 Nov '23

24 Nov '23

hulk inclusion category: performance bugzilla: https://gitee.com/openeuler/kernel/issues/I8J8UH -------------------------------- In the clock_read_data structure, the epoch_ns and epoch_cyc members are often updated by the update_sched_clock(). In fact, the sched_clock_mask, read_sched_clock, mult, and shift will no longer change after they are initialized and registered in sched_clock_register(). Because all of them are in the same cache line, especially in multi-core scenarios, this will cause frequent cache miss. Here we move these almostly read-only members into the struct clock_data to avoid false share with epoch_* members, and separate them into different cache line with padding. Here are the detailed test results of unixbench. ./Run -i 3 -c 96 context1 Without Patch ------------------------------------------------------------------------ Pipe-based Context Switching 7423119.7 lps (10.0 s, 2 samples) System Benchmarks Partial Index BASELINE RESULT INDEX Pipe-based Context Switching 4000.0 7423119.7 18557.8 ======== System Benchmarks Index Score (Partial Only) 18557.8 With Patch ------------------------------------------------------------------------ Pipe-based Context Switching 15869071.9 lps (10.0 s, 2 samples) System Benchmarks Partial Index BASELINE RESULT INDEX Pipe-based Context Switching 4000.0 15869071.9 39672.7 ======== System Benchmarks Index Score (Partial Only) 39672.7 Signed-off-by: Zeng Heng <zengheng4(a)huawei.com> --- arch/arm64/kernel/perf_event.c | 12 ++++---- include/linux/sched_clock.h | 42 ++++++++++++++++++++++---- kernel/time/sched_clock.c | 55 ++++++++++------------------------ 3 files changed, 59 insertions(+), 50 deletions(-) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index cdb3d4549b3a..4894fb0668fa 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -1255,6 +1255,8 @@ static int __init armv8_pmu_driver_init(void) } device_initcall(armv8_pmu_driver_init) +extern struct clock_data cd; + void arch_perf_update_userpage(struct perf_event *event, struct perf_event_mmap_page *userpg, u64 now) { @@ -1269,21 +1271,21 @@ void arch_perf_update_userpage(struct perf_event *event, do { rd = sched_clock_read_begin(&seq); - if (rd->read_sched_clock != arch_timer_read_counter) + if (cd.read_sched_clock != arch_timer_read_counter) return; - userpg->time_mult = rd->mult; - userpg->time_shift = rd->shift; + userpg->time_mult = cd.mult; + userpg->time_shift = cd.shift; userpg->time_zero = rd->epoch_ns; userpg->time_cycles = rd->epoch_cyc; - userpg->time_mask = rd->sched_clock_mask; + userpg->time_mask = cd.sched_clock_mask; /* * Subtract the cycle base, such that software that * doesn't know about cap_user_time_short still 'works' * assuming no wraps. */ - ns = mul_u64_u32_shr(rd->epoch_cyc, rd->mult, rd->shift); + ns = mul_u64_u32_shr(rd->epoch_cyc, cd.mult, cd.shift); userpg->time_zero -= ns; } while (sched_clock_read_retry(seq)); diff --git a/include/linux/sched_clock.h b/include/linux/sched_clock.h index 835ee87ed792..579b6d02c80c 100644 --- a/include/linux/sched_clock.h +++ b/include/linux/sched_clock.h @@ -11,11 +11,6 @@ * * @epoch_ns: sched_clock() value at last update * @epoch_cyc: Clock cycle value at last update. - * @sched_clock_mask: Bitmask for two's complement subtraction of non 64bit - * clocks. - * @read_sched_clock: Current clock source (or dummy source when suspended). - * @mult: Multiplier for scaled math conversion. - * @shift: Shift value for scaled math conversion. * * Care must be taken when updating this structure; it is read by * some very hot code paths. It occupies <=40 bytes and, when combined @@ -25,10 +20,47 @@ struct clock_read_data { u64 epoch_ns; u64 epoch_cyc; +}; + +/** + * struct clock_data - all data needed for sched_clock() (including + * registration of a new clock source) + * + * @seq: Sequence counter for protecting updates. The lowest + * bit is the index for @read_data. + * @read_data: Data required to read from sched_clock. + * + * @sched_clock_mask: Bitmask for two's complement subtraction of non 64bit + * clocks. + * @read_sched_clock: Current clock source (or dummy source when suspended). + * @mult: Multiplier for scaled math conversion. + * @shift: Shift value for scaled math conversion. + * + * @wrap_kt: Duration for which clock can run before wrapping. + * @rate: Tick rate of the registered clock. + * @actual_read_sched_clock: Registered hardware level clock read function. + * + * The ordering of this structure has been chosen to optimize cache + * performance. In particular 'seq' and 'read_data[0]' (combined) should fit + * into a single 64-byte cache line. + */ +struct clock_data { + seqcount_latch_t seq; + struct clock_read_data read_data[2]; + + struct { + char padding[0]; + } ____cacheline_aligned; + u64 sched_clock_mask; u64 (*read_sched_clock)(void); u32 mult; u32 shift; + + ktime_t wrap_kt; + unsigned long rate; + + u64 (*actual_read_sched_clock)(void); }; extern struct clock_read_data *sched_clock_read_begin(unsigned int *seq); diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c index b1b9b12899f5..de50a321f835 100644 --- a/kernel/time/sched_clock.c +++ b/kernel/time/sched_clock.c @@ -19,30 +19,6 @@ #include "timekeeping.h" -/** - * struct clock_data - all data needed for sched_clock() (including - * registration of a new clock source) - * - * @seq: Sequence counter for protecting updates. The lowest - * bit is the index for @read_data. - * @read_data: Data required to read from sched_clock. - * @wrap_kt: Duration for which clock can run before wrapping. - * @rate: Tick rate of the registered clock. - * @actual_read_sched_clock: Registered hardware level clock read function. - * - * The ordering of this structure has been chosen to optimize cache - * performance. In particular 'seq' and 'read_data[0]' (combined) should fit - * into a single 64-byte cache line. - */ -struct clock_data { - seqcount_latch_t seq; - struct clock_read_data read_data[2]; - ktime_t wrap_kt; - unsigned long rate; - - u64 (*actual_read_sched_clock)(void); -}; - static struct hrtimer sched_clock_timer; static int irqtime = -1; @@ -57,9 +33,9 @@ static u64 notrace jiffy_sched_clock_read(void) return (u64)(jiffies - INITIAL_JIFFIES); } -static struct clock_data cd ____cacheline_aligned = { - .read_data[0] = { .mult = NSEC_PER_SEC / HZ, - .read_sched_clock = jiffy_sched_clock_read, }, +struct clock_data cd ____cacheline_aligned = { + .mult = NSEC_PER_SEC / HZ, + .read_sched_clock = jiffy_sched_clock_read, .actual_read_sched_clock = jiffy_sched_clock_read, }; @@ -88,9 +64,9 @@ unsigned long long notrace sched_clock(void) do { rd = sched_clock_read_begin(&seq); - cyc = (rd->read_sched_clock() - rd->epoch_cyc) & - rd->sched_clock_mask; - res = rd->epoch_ns + cyc_to_ns(cyc, rd->mult, rd->shift); + cyc = (cd.read_sched_clock() - rd->epoch_cyc) & + cd.sched_clock_mask; + res = rd->epoch_ns + cyc_to_ns(cyc, cd.mult, cd.shift); } while (sched_clock_read_retry(seq)); return res; @@ -133,7 +109,7 @@ static void update_sched_clock(void) rd = cd.read_data[0]; cyc = cd.actual_read_sched_clock(); - ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & rd.sched_clock_mask, rd.mult, rd.shift); + ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & cd.sched_clock_mask, cd.mult, cd.shift); rd.epoch_ns = ns; rd.epoch_cyc = cyc; @@ -179,13 +155,14 @@ sched_clock_register(u64 (*read)(void), int bits, unsigned long rate) /* Update epoch for new counter and update 'epoch_ns' from old counter*/ new_epoch = read(); cyc = cd.actual_read_sched_clock(); - ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & rd.sched_clock_mask, rd.mult, rd.shift); + ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & cd.sched_clock_mask, cd.mult, cd.shift); + cd.actual_read_sched_clock = read; + cd.read_sched_clock = read; + cd.sched_clock_mask = new_mask; + cd.mult = new_mult; + cd.shift = new_shift; - rd.read_sched_clock = read; - rd.sched_clock_mask = new_mask; - rd.mult = new_mult; - rd.shift = new_shift; rd.epoch_cyc = new_epoch; rd.epoch_ns = ns; @@ -265,11 +242,9 @@ static u64 notrace suspended_sched_clock_read(void) int sched_clock_suspend(void) { - struct clock_read_data *rd = &cd.read_data[0]; - update_sched_clock(); hrtimer_cancel(&sched_clock_timer); - rd->read_sched_clock = suspended_sched_clock_read; + cd.read_sched_clock = suspended_sched_clock_read; return 0; } @@ -280,7 +255,7 @@ void sched_clock_resume(void) rd->epoch_cyc = cd.actual_read_sched_clock(); hrtimer_start(&sched_clock_timer, cd.wrap_kt, HRTIMER_MODE_REL_HARD); - rd->read_sched_clock = cd.actual_read_sched_clock; + cd.read_sched_clock = cd.actual_read_sched_clock; } static struct syscore_ops sched_clock_ops = { -- 2.25.1

2 1

[PATCH OLK-5.10] time/sched_clock: Remove clock_read_data's read mostly members
by Zeng Heng 24 Nov '23

24 Nov '23

In the clock_read_data structure, the epoch_ns and epoch_cyc members are often updated by the update_sched_clock(). In fact, the sched_clock_mask, read_sched_clock, mult, and shift will no longer change after they are initialized and registered in sched_clock_register(). Because all of them are in the same cache line, especially in multi-core scenarios, this will cause frequent cache miss. Here we move these almostly read-only members into the struct clock_data to avoid false share with epoch_* members, and separate them into different cache line with padding. Here are the detailed test results of unixbench. ./Run -i 3 -c 96 context1 Without Patch ------------------------------------------------------------------------ Pipe-based Context Switching 7423119.7 lps (10.0 s, 2 samples) System Benchmarks Partial Index BASELINE RESULT INDEX Pipe-based Context Switching 4000.0 7423119.7 18557.8 ======== System Benchmarks Index Score (Partial Only) 18557.8 With Patch ------------------------------------------------------------------------ Pipe-based Context Switching 15869071.9 lps (10.0 s, 2 samples) System Benchmarks Partial Index BASELINE RESULT INDEX Pipe-based Context Switching 4000.0 15869071.9 39672.7 ======== System Benchmarks Index Score (Partial Only) 39672.7 Signed-off-by: Zeng Heng <zengheng4(a)huawei.com> --- arch/arm64/kernel/perf_event.c | 12 ++++---- include/linux/sched_clock.h | 42 ++++++++++++++++++++++---- kernel/time/sched_clock.c | 55 ++++++++++------------------------ 3 files changed, 59 insertions(+), 50 deletions(-) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index cdb3d4549b3a..4894fb0668fa 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -1255,6 +1255,8 @@ static int __init armv8_pmu_driver_init(void) } device_initcall(armv8_pmu_driver_init) +extern struct clock_data cd; + void arch_perf_update_userpage(struct perf_event *event, struct perf_event_mmap_page *userpg, u64 now) { @@ -1269,21 +1271,21 @@ void arch_perf_update_userpage(struct perf_event *event, do { rd = sched_clock_read_begin(&seq); - if (rd->read_sched_clock != arch_timer_read_counter) + if (cd.read_sched_clock != arch_timer_read_counter) return; - userpg->time_mult = rd->mult; - userpg->time_shift = rd->shift; + userpg->time_mult = cd.mult; + userpg->time_shift = cd.shift; userpg->time_zero = rd->epoch_ns; userpg->time_cycles = rd->epoch_cyc; - userpg->time_mask = rd->sched_clock_mask; + userpg->time_mask = cd.sched_clock_mask; /* * Subtract the cycle base, such that software that * doesn't know about cap_user_time_short still 'works' * assuming no wraps. */ - ns = mul_u64_u32_shr(rd->epoch_cyc, rd->mult, rd->shift); + ns = mul_u64_u32_shr(rd->epoch_cyc, cd.mult, cd.shift); userpg->time_zero -= ns; } while (sched_clock_read_retry(seq)); diff --git a/include/linux/sched_clock.h b/include/linux/sched_clock.h index 835ee87ed792..579b6d02c80c 100644 --- a/include/linux/sched_clock.h +++ b/include/linux/sched_clock.h @@ -11,11 +11,6 @@ * * @epoch_ns: sched_clock() value at last update * @epoch_cyc: Clock cycle value at last update. - * @sched_clock_mask: Bitmask for two's complement subtraction of non 64bit - * clocks. - * @read_sched_clock: Current clock source (or dummy source when suspended). - * @mult: Multiplier for scaled math conversion. - * @shift: Shift value for scaled math conversion. * * Care must be taken when updating this structure; it is read by * some very hot code paths. It occupies <=40 bytes and, when combined @@ -25,10 +20,47 @@ struct clock_read_data { u64 epoch_ns; u64 epoch_cyc; +}; + +/** + * struct clock_data - all data needed for sched_clock() (including + * registration of a new clock source) + * + * @seq: Sequence counter for protecting updates. The lowest + * bit is the index for @read_data. + * @read_data: Data required to read from sched_clock. + * + * @sched_clock_mask: Bitmask for two's complement subtraction of non 64bit + * clocks. + * @read_sched_clock: Current clock source (or dummy source when suspended). + * @mult: Multiplier for scaled math conversion. + * @shift: Shift value for scaled math conversion. + * + * @wrap_kt: Duration for which clock can run before wrapping. + * @rate: Tick rate of the registered clock. + * @actual_read_sched_clock: Registered hardware level clock read function. + * + * The ordering of this structure has been chosen to optimize cache + * performance. In particular 'seq' and 'read_data[0]' (combined) should fit + * into a single 64-byte cache line. + */ +struct clock_data { + seqcount_latch_t seq; + struct clock_read_data read_data[2]; + + struct { + char padding[0]; + } ____cacheline_aligned; + u64 sched_clock_mask; u64 (*read_sched_clock)(void); u32 mult; u32 shift; + + ktime_t wrap_kt; + unsigned long rate; + + u64 (*actual_read_sched_clock)(void); }; extern struct clock_read_data *sched_clock_read_begin(unsigned int *seq); diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c index b1b9b12899f5..de50a321f835 100644 --- a/kernel/time/sched_clock.c +++ b/kernel/time/sched_clock.c @@ -19,30 +19,6 @@ #include "timekeeping.h" -/** - * struct clock_data - all data needed for sched_clock() (including - * registration of a new clock source) - * - * @seq: Sequence counter for protecting updates. The lowest - * bit is the index for @read_data. - * @read_data: Data required to read from sched_clock. - * @wrap_kt: Duration for which clock can run before wrapping. - * @rate: Tick rate of the registered clock. - * @actual_read_sched_clock: Registered hardware level clock read function. - * - * The ordering of this structure has been chosen to optimize cache - * performance. In particular 'seq' and 'read_data[0]' (combined) should fit - * into a single 64-byte cache line. - */ -struct clock_data { - seqcount_latch_t seq; - struct clock_read_data read_data[2]; - ktime_t wrap_kt; - unsigned long rate; - - u64 (*actual_read_sched_clock)(void); -}; - static struct hrtimer sched_clock_timer; static int irqtime = -1; @@ -57,9 +33,9 @@ static u64 notrace jiffy_sched_clock_read(void) return (u64)(jiffies - INITIAL_JIFFIES); } -static struct clock_data cd ____cacheline_aligned = { - .read_data[0] = { .mult = NSEC_PER_SEC / HZ, - .read_sched_clock = jiffy_sched_clock_read, }, +struct clock_data cd ____cacheline_aligned = { + .mult = NSEC_PER_SEC / HZ, + .read_sched_clock = jiffy_sched_clock_read, .actual_read_sched_clock = jiffy_sched_clock_read, }; @@ -88,9 +64,9 @@ unsigned long long notrace sched_clock(void) do { rd = sched_clock_read_begin(&seq); - cyc = (rd->read_sched_clock() - rd->epoch_cyc) & - rd->sched_clock_mask; - res = rd->epoch_ns + cyc_to_ns(cyc, rd->mult, rd->shift); + cyc = (cd.read_sched_clock() - rd->epoch_cyc) & + cd.sched_clock_mask; + res = rd->epoch_ns + cyc_to_ns(cyc, cd.mult, cd.shift); } while (sched_clock_read_retry(seq)); return res; @@ -133,7 +109,7 @@ static void update_sched_clock(void) rd = cd.read_data[0]; cyc = cd.actual_read_sched_clock(); - ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & rd.sched_clock_mask, rd.mult, rd.shift); + ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & cd.sched_clock_mask, cd.mult, cd.shift); rd.epoch_ns = ns; rd.epoch_cyc = cyc; @@ -179,13 +155,14 @@ sched_clock_register(u64 (*read)(void), int bits, unsigned long rate) /* Update epoch for new counter and update 'epoch_ns' from old counter*/ new_epoch = read(); cyc = cd.actual_read_sched_clock(); - ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & rd.sched_clock_mask, rd.mult, rd.shift); + ns = rd.epoch_ns + cyc_to_ns((cyc - rd.epoch_cyc) & cd.sched_clock_mask, cd.mult, cd.shift); + cd.actual_read_sched_clock = read; + cd.read_sched_clock = read; + cd.sched_clock_mask = new_mask; + cd.mult = new_mult; + cd.shift = new_shift; - rd.read_sched_clock = read; - rd.sched_clock_mask = new_mask; - rd.mult = new_mult; - rd.shift = new_shift; rd.epoch_cyc = new_epoch; rd.epoch_ns = ns; @@ -265,11 +242,9 @@ static u64 notrace suspended_sched_clock_read(void) int sched_clock_suspend(void) { - struct clock_read_data *rd = &cd.read_data[0]; - update_sched_clock(); hrtimer_cancel(&sched_clock_timer); - rd->read_sched_clock = suspended_sched_clock_read; + cd.read_sched_clock = suspended_sched_clock_read; return 0; } @@ -280,7 +255,7 @@ void sched_clock_resume(void) rd->epoch_cyc = cd.actual_read_sched_clock(); hrtimer_start(&sched_clock_timer, cd.wrap_kt, HRTIMER_MODE_REL_HARD); - rd->read_sched_clock = cd.actual_read_sched_clock; + cd.read_sched_clock = cd.actual_read_sched_clock; } static struct syscore_ops sched_clock_ops = { -- 2.25.1

2 1