- Kernel - mailweb.openeuler.org

[PATCH openEuler-1.0-LTS 1/2] mm: share_pool: adjust sp_make_share_k2u behavior when coredump
by Yang Yingliang 16 Dec '21

16 Dec '21

From: Guo Mengqi <guomengqi3(a)huawei.com> ascend inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I4MUV2 CVE: NA when k2u is being executed ont the whole sharepool group, and one process coredumps, k2u will skip the coredumped process and continue on the rest processes in the group. Signed-off-by: Guo Mengqi <guomengqi3(a)huawei.com> Reviewed-by: Weilong Chen <chenweilong(a)huawei.com> Signed-off-by: Yang Yingliang <yangyingliang(a)huawei.com> --- mm/share_pool.c | 50 +++++++++++++++++++++++++++++++------------------ 1 file changed, 32 insertions(+), 18 deletions(-) diff --git a/mm/share_pool.c b/mm/share_pool.c index c9079e6a40b1a..ac97a68a70264 100644 --- a/mm/share_pool.c +++ b/mm/share_pool.c @@ -678,8 +678,25 @@ static unsigned long sp_mmap(struct mm_struct *mm, struct file *file, struct sp_area *spa, unsigned long *populate, unsigned long prot); static void sp_munmap(struct mm_struct *mm, unsigned long addr, unsigned long size); + +#define K2U_NORMAL 0 +#define K2U_COREDUMP 1 + +struct sp_k2u_context { + unsigned long kva; + unsigned long kva_aligned; + unsigned long size; + unsigned long size_aligned; + unsigned long sp_flags; + int state; + int spg_id; + bool to_task; + struct timespec64 start; + struct timespec64 end; +}; + static unsigned long sp_remap_kva_to_vma(unsigned long kva, struct sp_area *spa, - struct mm_struct *mm, unsigned long prot); + struct mm_struct *mm, unsigned long prot, struct sp_k2u_context *kc); static void free_sp_group_id(int spg_id) { @@ -1334,7 +1351,7 @@ int mg_sp_group_add_task(int pid, unsigned long prot, int spg_id) spin_unlock(&sp_area_lock); if (spa->type == SPA_TYPE_K2SPG && spa->kva) { - addr = sp_remap_kva_to_vma(spa->kva, spa, mm, prot); + addr = sp_remap_kva_to_vma(spa->kva, spa, mm, prot, NULL); if (IS_ERR_VALUE(addr)) pr_warn("add group remap k2u failed %ld\n", addr); @@ -2606,7 +2623,7 @@ static unsigned long __sp_remap_get_pfn(unsigned long kva) /* when called by k2u to group, always make sure rw_lock of spg is down */ static unsigned long sp_remap_kva_to_vma(unsigned long kva, struct sp_area *spa, - struct mm_struct *mm, unsigned long prot) + struct mm_struct *mm, unsigned long prot, struct sp_k2u_context *kc) { struct vm_area_struct *vma; unsigned long ret_addr; @@ -2618,6 +2635,8 @@ static unsigned long sp_remap_kva_to_vma(unsigned long kva, struct sp_area *spa, if (unlikely(mm->core_state)) { pr_err("k2u mmap: encountered coredump, abort\n"); ret_addr = -EBUSY; + if (kc) + kc->state = K2U_COREDUMP; goto put_mm; } @@ -2703,7 +2722,7 @@ static void *sp_make_share_kva_to_task(unsigned long kva, unsigned long size, un spa->kva = kva; - uva = (void *)sp_remap_kva_to_vma(kva, spa, current->mm, prot); + uva = (void *)sp_remap_kva_to_vma(kva, spa, current->mm, prot, NULL); __sp_area_drop(spa); if (IS_ERR(uva)) pr_err("remap k2u to task failed %ld\n", PTR_ERR(uva)); @@ -2731,6 +2750,8 @@ static void *sp_make_share_kva_to_spg(unsigned long kva, unsigned long size, struct mm_struct *mm; struct sp_group_node *spg_node; void *uva = ERR_PTR(-ENODEV); + struct sp_k2u_context kc; + unsigned long ret_addr = -ENODEV; down_read(&spg->rw_lock); spa = sp_alloc_area(size, sp_flags, spg, SPA_TYPE_K2SPG, current->tgid); @@ -2745,12 +2766,17 @@ static void *sp_make_share_kva_to_spg(unsigned long kva, unsigned long size, list_for_each_entry(spg_node, &spg->procs, proc_node) { mm = spg_node->master->mm; - uva = (void *)sp_remap_kva_to_vma(kva, spa, mm, spg_node->prot); - if (IS_ERR(uva)) { + kc.state = K2U_NORMAL; + ret_addr = sp_remap_kva_to_vma(kva, spa, mm, spg_node->prot, &kc); + if (IS_ERR_VALUE(ret_addr)) { + if (kc.state == K2U_COREDUMP) + continue; + uva = (void *)ret_addr; pr_err("remap k2u to spg failed %ld\n", PTR_ERR(uva)); __sp_free(spg, spa->va_start, spa_size(spa), mm); goto out; } + uva = (void *)ret_addr; } out: @@ -2775,18 +2801,6 @@ static bool vmalloc_area_set_flag(unsigned long kva, unsigned long flags) return false; } -struct sp_k2u_context { - unsigned long kva; - unsigned long kva_aligned; - unsigned long size; - unsigned long size_aligned; - unsigned long sp_flags; - int spg_id; - bool to_task; - struct timespec64 start; - struct timespec64 end; -}; - static void trace_sp_k2u_begin(struct sp_k2u_context *kc) { if (!sysctl_sp_perf_k2u) -- 2.25.1

1 1

[PATCH OLK-5.10 0/2]drm/hisilicon: Support hibmc get EDID
by Gou Hao 16 Dec '21

16 Dec '21

bugzilla: https://gitee.com/openeuler/kernel/issues/I469VQ Tian Tao (2): drm/hisilicon: Support i2c driver algorithms for bit-shift adapters drm/hisilicon: Features to support reading resolutions from EDID drivers/gpu/drm/hisilicon/hibmc/Makefile | 2 +- .../gpu/drm/hisilicon/hibmc/hibmc_drm_drv.h | 25 ++++- .../gpu/drm/hisilicon/hibmc/hibmc_drm_i2c.c | 99 +++++++++++++++++++ .../gpu/drm/hisilicon/hibmc/hibmc_drm_vdac.c | 38 ++++++- 4 files changed, 158 insertions(+), 6 deletions(-) create mode 100644 drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_i2c.c -- 2.20.1

1 2

[PATCH OLK-5.10 0/2]drm/hisilicon: Support hibmc get EDID
by Gou Hao 16 Dec '21

16 Dec '21

bugzilla: https://gitee.com/openeuler/kernel/issues/I469VQ Tian Tao (2): drm/hisilicon: Support i2c driver algorithms for bit-shift adapters drm/hisilicon: Features to support reading resolutions from EDID drivers/gpu/drm/hisilicon/hibmc/Makefile | 2 +- .../gpu/drm/hisilicon/hibmc/hibmc_drm_drv.h | 25 ++++- .../gpu/drm/hisilicon/hibmc/hibmc_drm_i2c.c | 99 +++++++++++++++++++ .../gpu/drm/hisilicon/hibmc/hibmc_drm_vdac.c | 38 ++++++- 4 files changed, 158 insertions(+), 6 deletions(-) create mode 100644 drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_i2c.c -- 2.20.1

1 1

[PATCH openEuler-1.0-LTS v10 00/26] support psi feature
by Liu Xinpeng 16 Dec '21

16 Dec '21

backport psi feature and avoid kabi change bugzilla: https://gitee.com/openeuler/kernel/issues/I47QS2 Changes since v9: 1.There are some problems when allocating cgroups, from Cheng Jian. First. We use `sizeof(struct cgroup_psi)` to allocate more space. This is inappropriate. Second, variable length arrays isn't the last member of cgroups_psi, so the offset of the psi will not be a compile-determined value. It's dangerous to use cgroup_psi(). 2.free path should change too,from Xie Xiuqi. Baruch Siach (1): psi: fix reference to kernel commandline enable Dan Schatzberg (1): kernel/sched/psi.c: expose pressure metrics on root cgroup Johannes Weiner (12): mm: workingset: tell cache transitions from workingset thrashing sched: loadavg: consolidate LOAD_INT, LOAD_FRAC, CALC_LOAD sched: loadavg: make calc_load_n() public sched: sched.h: make rq locking and clock functions available in stats.h sched: introduce this_rq_lock_irq() psi: pressure stall information for CPU, memory, and IO psi: cgroup support psi: make disabling/enabling easier for vendor kernels psi: fix aggregation idle shut-off psi: avoid divide-by-zero crash inside virtual machines fs: kernfs: add poll file operation sched/psi: Fix sampling error and rare div0 crashes with cgroups and high uptime Josef Bacik (1): blk-iolatency: use a percentile approache for ssd's Liu Xinpeng (2): psi:enable psi in config psi:avoid kabi change Miklos Szeredi (1): fuse: ignore PG_workingset after stealing Olof Johansson (1): kernel/sched/psi.c: simplify cgroup_move_task() Suren Baghdasaryan (6): psi: introduce state_mask to represent stalled psi states psi: make psi_enable static psi: rename psi fields in preparation for psi trigger addition psi: split update_stats into parts psi: track changed states include/: refactor headers to allow kthread.h inclusion in psi_types.h Yafang Shao (1): mm, memcg: add workingset_restore in memory.stat Documentation/accounting/psi.txt | 73 +++ Documentation/admin-guide/cgroup-v2.rst | 22 + Documentation/admin-guide/kernel-parameters.txt | 4 + arch/arm64/configs/openeuler_defconfig | 2 + arch/powerpc/platforms/cell/cpufreq_spudemand.c | 2 +- arch/powerpc/platforms/cell/spufs/sched.c | 9 +- arch/s390/appldata/appldata_os.c | 4 - arch/x86/configs/openeuler_defconfig | 2 + block/blk-iolatency.c | 183 +++++- drivers/cpuidle/governors/menu.c | 4 - drivers/spi/spi-rockchip.c | 1 + fs/fuse/dev.c | 1 + fs/kernfs/file.c | 31 +- fs/proc/loadavg.c | 3 - include/linux/cgroup-defs.h | 12 + include/linux/cgroup.h | 17 + include/linux/kernfs.h | 8 + include/linux/kthread.h | 4 + include/linux/page-flags.h | 5 + include/linux/psi.h | 55 ++ include/linux/psi_types.h | 95 +++ include/linux/sched.h | 13 + include/linux/sched/loadavg.h | 24 +- include/linux/swap.h | 1 + include/trace/events/mmflags.h | 1 + init/Kconfig | 28 + kernel/cgroup/cgroup.c | 69 +- kernel/debug/kdb/kdb_main.c | 7 +- kernel/fork.c | 4 + kernel/kthread.c | 3 + kernel/sched/Makefile | 1 + kernel/sched/core.c | 16 +- kernel/sched/loadavg.c | 139 ++-- kernel/sched/psi.c | 823 ++++++++++++++++++++++++ kernel/sched/sched.h | 178 ++--- kernel/sched/stats.h | 86 +++ kernel/workqueue.c | 23 + kernel/workqueue_internal.h | 6 +- mm/compaction.c | 5 + mm/filemap.c | 20 +- mm/huge_memory.c | 1 + mm/migrate.c | 2 + mm/page_alloc.c | 9 + mm/swap_state.c | 1 + mm/vmscan.c | 10 + mm/workingset.c | 113 +++- 46 files changed, 1841 insertions(+), 279 deletions(-) create mode 100644 Documentation/accounting/psi.txt create mode 100644 include/linux/psi.h create mode 100644 include/linux/psi_types.h create mode 100644 kernel/sched/psi.c -- 1.8.3.1

3 28

[PATCH openEuler-1.0-LTS 1/4] time: Normalize timespec64 before timespec64_compare()
by Yang Yingliang 15 Dec '21

15 Dec '21

From: Yu Liao <liaoyu15(a)huawei.com> hulk inclusion category: bugfix bugzilla: 185831, https://gitee.com/openeuler/kernel/issues/I4MO2G CVE: NA ------------------------------------------------- Passing unnormalized timespec64 to timespec64_compare() may cause incorrect results. For example: wall_to_monotonic = {tv_sec = -10, tv_nsec = 900000000} ts_delta = {tv_sec = -9, tv_nsec = -900000000} timespec64_compare() returns -1, but actually wall_to_monotonic > ts_delta. This will cause wall_to_monotonic to become a positive number. Use timespec64_sub() instead of direct subtraction to avoid this. Signed-off-by: Yu Liao <liaoyu15(a)huawei.com> Reviewed-by: Xiongfeng Wang <wangxiongfeng2(a)huawei.com> Signed-off-by: Yang Yingliang <yangyingliang(a)huawei.com> --- include/linux/time64.h | 2 ++ kernel/time/timekeeping.c | 3 +-- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/include/linux/time64.h b/include/linux/time64.h index b58142aa0b1dd..e58f59674ad6d 100644 --- a/include/linux/time64.h +++ b/include/linux/time64.h @@ -62,6 +62,8 @@ static inline int timespec64_equal(const struct timespec64 *a, * lhs < rhs: return <0 * lhs == rhs: return 0 * lhs > rhs: return >0 + * + * Note: Both lhs and rhs must be normalized. */ static inline int timespec64_compare(const struct timespec64 *lhs, const struct timespec64 *rhs) { diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 64aa492f79422..17ba0a2f5f5d1 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -1236,8 +1236,7 @@ int do_settimeofday64(const struct timespec64 *ts) timekeeping_forward_now(tk); xt = tk_xtime(tk); - ts_delta.tv_sec = ts->tv_sec - xt.tv_sec; - ts_delta.tv_nsec = ts->tv_nsec - xt.tv_nsec; + ts_delta = timespec64_sub(*ts, xt); if (timespec64_compare(&tk->wall_to_monotonic, &ts_delta) > 0) { ret = -EINVAL; -- 2.25.1

1 3

[PATCH] scsi: pm8001: Fix memory leak in pm8001_send_abort_all()
by Jianglei Nie 15 Dec '21

15 Dec '21

openEuler inclusion category: bugfix bugzilla: NA CVE: NA ------------------------ In line 1767, sas_alloc_slow_task() allocates and initializes a sas_task structure. When some errors occur, line 1778 and line 1795 forget to free this structure, which will lead to a memory leak. There is a similar snippet of code in the same file (in function pm8001_send_read_log) as allocating and initializing in line 1812 as well as releasing the memory in line 1822 and line 1867. We can fix it by calling sas_free_task() when the res and ret is true and before the function returns. Signed-off-by: Jianglei Nie <niejianglei2021(a)163.com> --- drivers/scsi/pm8001/pm8001_hwi.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/pm8001/pm8001_hwi.c b/drivers/scsi/pm8001/pm8001_hwi.c index 124cb69740c6..25045a91620e 100644 --- a/drivers/scsi/pm8001/pm8001_hwi.c +++ b/drivers/scsi/pm8001/pm8001_hwi.c @@ -1774,8 +1774,10 @@ static void pm8001_send_abort_all(struct pm8001_hba_info *pm8001_ha, task->task_done = pm8001_task_done; res = pm8001_tag_alloc(pm8001_ha, &ccb_tag); - if (res) + if (res) { + sas_free_task(task); return; + } ccb = &pm8001_ha->ccb_info[ccb_tag]; ccb->device = pm8001_ha_dev; @@ -1791,8 +1793,10 @@ static void pm8001_send_abort_all(struct pm8001_hba_info *pm8001_ha, ret = pm8001_mpi_build_cmd(pm8001_ha, circularQ, opc, &task_abort, sizeof(task_abort), 0); - if (ret) + if (ret) { + sas_free_task(task); pm8001_tag_free(pm8001_ha, ccb_tag); + } } -- 2.25.1

1 0

perf spe-c2c record失败的问题
by haoxin 15 Dec '21

15 Dec '21

您好，我采用最细的openEuler-1.0测试perf spe-c2c的功能，但是发现/sys/devices下没有arm_spe_0这个目录执行perf spe-c2c record时显示： event syntax error: 'arm_spe_0/ts_enable=1,pct_enable=1,pa_enable=1,load_filter=1,jitter=1,store_filter=1,min_latency=0/' \___ Cannot find PMU `arm_spe_0'. Missing kernel support? 想问下您这是为什么？需要哪里开启spe功能吗？

3 2

[PATCH openEuler-1.0-LTS] fix kabi effect by change in md_rdev
by Yang Yingliang 14 Dec '21

14 Dec '21

From: zhangwensheng <zhangwensheng5(a)huawei.com> hulk inclusion category: bugfix bugzilla: 185891 CVE: NA -------------------------------- because enlarge the last_events in md_rdev, to fix the kabi change; Signed-off-by: zhangwensheng <zhangwensheng5(a)huawei.com> Reviewed-by: Zhang Yi <yi.zhang(a)huawei.com> Signed-off-by: Yang Yingliang <yangyingliang(a)huawei.com> --- drivers/md/md.h | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/drivers/md/md.h b/drivers/md/md.h index e745f701e1a82..422af63f1e1ee 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -47,8 +47,16 @@ struct md_rdev { sector_t sectors; /* Device size (in 512bytes sectors) */ struct mddev *mddev; /* RAID array if running */ - long long last_events; /* IO event timestamp */ + /* + * IO event timestamp, this pos has 64 bit space, + * can enlarge it solving its small size. + */ +#ifndef __GENKSYMS__ + long long last_events; +#else + int last_events; +#endif /* * If meta_bdev is non-NULL, it means that a separate device is * being used to store the metadata (superblock/bitmap) which -- 2.25.1

1 0

[PATCH openEuler-1.0-LTS] Revert "dm space maps: don't reset space map allocation cursor when committing"
by Yang Yingliang 14 Dec '21

14 Dec '21

From: Luo Meng <luomeng12(a)huawei.com> hulk inclusion category: bugfix bugzilla: 185894, https://gitee.com/openeuler/kernel/issues/I4MITL CVE: NA ----------------------------------------------- This reverts commit 38cb5d45845b9d8962eb26660c14b7f05abc602f. Signed-off-by: Luo Meng <luomeng12(a)huawei.com> Reviewed-by: Jason Yan <yanaijie(a)huawei.com> Signed-off-by: Yang Yingliang <yangyingliang(a)huawei.com> --- drivers/md/persistent-data/dm-space-map-disk.c | 9 +-------- drivers/md/persistent-data/dm-space-map-metadata.c | 9 +-------- 2 files changed, 2 insertions(+), 16 deletions(-) diff --git a/drivers/md/persistent-data/dm-space-map-disk.c b/drivers/md/persistent-data/dm-space-map-disk.c index e0acae7a3815d..bf4c5e2ccb6ff 100644 --- a/drivers/md/persistent-data/dm-space-map-disk.c +++ b/drivers/md/persistent-data/dm-space-map-disk.c @@ -171,14 +171,6 @@ static int sm_disk_new_block(struct dm_space_map *sm, dm_block_t *b) * Any block we allocate has to be free in both the old and current ll. */ r = sm_ll_find_common_free_block(&smd->old_ll, &smd->ll, smd->begin, smd->ll.nr_blocks, b); - if (r == -ENOSPC) { - /* - * There's no free block between smd->begin and the end of the metadata device. - * We search before smd->begin in case something has been freed. - */ - r = sm_ll_find_common_free_block(&smd->old_ll, &smd->ll, 0, smd->begin, b); - } - if (r) return r; @@ -207,6 +199,7 @@ static int sm_disk_commit(struct dm_space_map *sm) return r; memcpy(&smd->old_ll, &smd->ll, sizeof(smd->old_ll)); + smd->begin = 0; smd->nr_allocated_this_transaction = 0; r = sm_disk_get_nr_free(sm, &nr_free); diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c index da439ac857963..9e3c64ec2026f 100644 --- a/drivers/md/persistent-data/dm-space-map-metadata.c +++ b/drivers/md/persistent-data/dm-space-map-metadata.c @@ -452,14 +452,6 @@ static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b) * Any block we allocate has to be free in both the old and current ll. */ r = sm_ll_find_common_free_block(&smm->old_ll, &smm->ll, smm->begin, smm->ll.nr_blocks, b); - if (r == -ENOSPC) { - /* - * There's no free block between smm->begin and the end of the metadata device. - * We search before smm->begin in case something has been freed. - */ - r = sm_ll_find_common_free_block(&smm->old_ll, &smm->ll, 0, smm->begin, b); - } - if (r) return r; @@ -511,6 +503,7 @@ static int sm_metadata_commit(struct dm_space_map *sm) return r; memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll)); + smm->begin = 0; smm->allocated_this_transaction = 0; return 0; -- 2.25.1

1 0

[PATCH openEuler-1.0-LTS 1/4] nvme-fabrics: reject I/O to offline device
by Yang Yingliang 14 Dec '21

14 Dec '21

From: Victor Gladkov <victor.gladkov(a)kioxia.com> mainline inclusion from mainline-v5.11-rc1 commit 8c4dfea97f15b80097b3f882ca428fb2751ec30c category: bugfix bugzilla: NA CVE: NA Link: https://gitee.com/openeuler/kernel/issues/I4JFPM?from=project-issue ------------------------------------------------- Commands get stuck while Host NVMe-oF controller is in reconnect state. The controller enters into reconnect state when it loses connection with the target. It tries to reconnect every 10 seconds (default) until a successful reconnect or until the reconnect time-out is reached. The default reconnect time out is 10 minutes. Applications are expecting commands to complete with success or error within a certain timeout (30 seconds by default). The NVMe host is enforcing that timeout while it is connected, but during reconnect the timeout is not enforced and commands may get stuck for a long period or even forever. To fix this long delay due to the default timeout, introduce new "fast_io_fail_tmo" session parameter. The timeout is measured in seconds from the controller reconnect and any command beyond that timeout is rejected. The new parameter value may be passed during 'connect'. The default value of -1 means no timeout (similar to current behavior). Signed-off-by: Victor Gladkov <victor.gladkov(a)kioxia.com> Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni(a)wdc.com> Reviewed-by: Hannes Reinecke <hare(a)suse.de> Reviewed-by: Sagi Grimberg <sagi(a)grimberg.me> Reviewed-by: Chao Leng <lengchao(a)huawei.com> Signed-off-by: Christoph Hellwig <hch(a)lst.de> conflicts:: drivers/nvme/host/core.c [adjust context] Signed-off-by: jiangtao <jiangtao62(a)huawei.com> Reviewed-by: chengjike <chengjike.cheng(a)huawei.com> Reviewed-by: Ao Sun <sunao.sun(a)huawei.com> Reviewed-by: Zhenwei Yang <yangzhenwei(a)huawei.com> Reviewed-by: Hou Tao <houtao1(a)huawei.com> Signed-off-by: Yang Yingliang <yangyingliang(a)huawei.com> --- drivers/nvme/host/core.c | 51 +++++++++++++++++++++++++++++++++-- drivers/nvme/host/fabrics.c | 28 ++++++++++++++++--- drivers/nvme/host/fabrics.h | 5 ++++ drivers/nvme/host/multipath.c | 2 ++ drivers/nvme/host/nvme.h | 3 +++ 5 files changed, 83 insertions(+), 6 deletions(-) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index b45ae13aa5fe6..c9ca375a5ddb9 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -130,6 +130,37 @@ static void nvme_queue_scan(struct nvme_ctrl *ctrl) queue_work(nvme_wq, &ctrl->scan_work); } +static void nvme_failfast_work(struct work_struct *work) +{ + struct nvme_ctrl *ctrl = container_of(to_delayed_work(work), + struct nvme_ctrl, failfast_work); + + if (ctrl->state != NVME_CTRL_CONNECTING) + return; + + set_bit(NVME_CTRL_FAILFAST_EXPIRED, &ctrl->flags); + dev_info(ctrl->device, "failfast expired\n"); + nvme_kick_requeue_lists(ctrl); +} + +static inline void nvme_start_failfast_work(struct nvme_ctrl *ctrl) +{ + if (!ctrl->opts || ctrl->opts->fast_io_fail_tmo == -1) + return; + + schedule_delayed_work(&ctrl->failfast_work, + ctrl->opts->fast_io_fail_tmo * HZ); +} + +static inline void nvme_stop_failfast_work(struct nvme_ctrl *ctrl) +{ + if (!ctrl->opts) + return; + + cancel_delayed_work_sync(&ctrl->failfast_work); + clear_bit(NVME_CTRL_FAILFAST_EXPIRED, &ctrl->flags); +} + int nvme_reset_ctrl(struct nvme_ctrl *ctrl) { if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_RESETTING)) @@ -385,8 +416,21 @@ bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl, ctrl->state = new_state; spin_unlock_irqrestore(&ctrl->lock, flags); - if (changed && ctrl->state == NVME_CTRL_LIVE) - nvme_kick_requeue_lists(ctrl); + if (changed) { + switch (ctrl->state) { + case NVME_CTRL_LIVE: + if (old_state == NVME_CTRL_CONNECTING) + nvme_stop_failfast_work(ctrl); + nvme_kick_requeue_lists(ctrl); + break; + case NVME_CTRL_CONNECTING: + if (old_state == NVME_CTRL_RESETTING) + nvme_start_failfast_work(ctrl); + break; + default: + break; + } + } return changed; } EXPORT_SYMBOL_GPL(nvme_change_ctrl_state); @@ -3712,6 +3756,7 @@ void nvme_stop_ctrl(struct nvme_ctrl *ctrl) { nvme_mpath_stop(ctrl); nvme_stop_keep_alive(ctrl); + nvme_stop_failfast_work(ctrl); flush_work(&ctrl->async_event_work); cancel_work_sync(&ctrl->fw_act_work); if (ctrl->ops->stop_ctrl) @@ -3776,6 +3821,7 @@ int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev, int ret; ctrl->state = NVME_CTRL_NEW; + clear_bit(NVME_CTRL_FAILFAST_EXPIRED, &ctrl->flags); spin_lock_init(&ctrl->lock); mutex_init(&ctrl->scan_lock); INIT_LIST_HEAD(&ctrl->namespaces); @@ -3791,6 +3837,7 @@ int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev, INIT_DELAYED_WORK(&ctrl->ka_work, nvme_keep_alive_work); memset(&ctrl->ka_cmd, 0, sizeof(ctrl->ka_cmd)); ctrl->ka_cmd.common.opcode = nvme_admin_keep_alive; + INIT_DELAYED_WORK(&ctrl->failfast_work, nvme_failfast_work); BUILD_BUG_ON(NVME_DSM_MAX_RANGES * sizeof(struct nvme_dsm_range) > PAGE_SIZE); diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c index 738794af3f389..650b3bd89968f 100644 --- a/drivers/nvme/host/fabrics.c +++ b/drivers/nvme/host/fabrics.c @@ -550,6 +550,7 @@ blk_status_t nvmf_fail_nonready_command(struct nvme_ctrl *ctrl, { if (ctrl->state != NVME_CTRL_DELETING && ctrl->state != NVME_CTRL_DEAD && + !test_bit(NVME_CTRL_FAILFAST_EXPIRED, &ctrl->flags) && !blk_noretry_request(rq) && !(rq->cmd_flags & REQ_NVME_MPATH)) return BLK_STS_RESOURCE; @@ -607,6 +608,7 @@ static const match_table_t opt_tokens = { { NVMF_OPT_HOST_TRADDR, "host_traddr=%s" }, { NVMF_OPT_HOST_ID, "hostid=%s" }, { NVMF_OPT_DUP_CONNECT, "duplicate_connect" }, + { NVMF_OPT_FAIL_FAST_TMO, "fast_io_fail_tmo=%d" }, { NVMF_OPT_ERR, NULL } }; @@ -626,6 +628,7 @@ static int nvmf_parse_options(struct nvmf_ctrl_options *opts, opts->reconnect_delay = NVMF_DEF_RECONNECT_DELAY; opts->kato = NVME_DEFAULT_KATO; opts->duplicate_connect = false; + opts->fast_io_fail_tmo = NVMF_DEF_FAIL_FAST_TMO; options = o = kstrdup(buf, GFP_KERNEL); if (!options) @@ -750,6 +753,17 @@ static int nvmf_parse_options(struct nvmf_ctrl_options *opts, pr_warn("ctrl_loss_tmo < 0 will reconnect forever\n"); ctrl_loss_tmo = token; break; + case NVMF_OPT_FAIL_FAST_TMO: + if (match_int(args, &token)) { + ret = -EINVAL; + goto out; + } + + if (token >= 0) + pr_warn("I/O will fail on after %d sec reconnect\n", + token); + opts->fast_io_fail_tmo = token; + break; case NVMF_OPT_HOSTNQN: if (opts->host) { pr_err("hostnqn already user-assigned: %s\n", @@ -830,11 +844,17 @@ static int nvmf_parse_options(struct nvmf_ctrl_options *opts, opts->nr_io_queues = 0; opts->duplicate_connect = true; } - if (ctrl_loss_tmo < 0) + + if (ctrl_loss_tmo < 0) { opts->max_reconnects = -1; - else + } else { opts->max_reconnects = DIV_ROUND_UP(ctrl_loss_tmo, opts->reconnect_delay); + if (ctrl_loss_tmo < opts->fast_io_fail_tmo) + pr_warn("failfast tmo (%d) > ctrl_loss_tmo (%d)\n", + opts->fast_io_fail_tmo, + ctrl_loss_tmo); + } if (!opts->host) { kref_get(&nvmf_default_host->ref); @@ -903,8 +923,8 @@ EXPORT_SYMBOL_GPL(nvmf_free_options); #define NVMF_REQUIRED_OPTS (NVMF_OPT_TRANSPORT | NVMF_OPT_NQN) #define NVMF_ALLOWED_OPTS (NVMF_OPT_QUEUE_SIZE | NVMF_OPT_NR_IO_QUEUES | \ NVMF_OPT_KATO | NVMF_OPT_HOSTNQN | \ - NVMF_OPT_HOST_ID | NVMF_OPT_DUP_CONNECT) - + NVMF_OPT_HOST_ID | NVMF_OPT_DUP_CONNECT |\ + NVMF_OPT_FAIL_FAST_TMO) static struct nvme_ctrl * nvmf_create_ctrl(struct device *dev, const char *buf, size_t count) { diff --git a/drivers/nvme/host/fabrics.h b/drivers/nvme/host/fabrics.h index 188ebbeec32ce..645ed75f22980 100644 --- a/drivers/nvme/host/fabrics.h +++ b/drivers/nvme/host/fabrics.h @@ -24,6 +24,8 @@ /* default to 600 seconds of reconnect attempts before giving up */ #define NVMF_DEF_CTRL_LOSS_TMO 600 #define NVMF_DEF_RECONNECT_FOREVER -1 +/* default is -1: the fail fast mechanism is disabled */ +#define NVMF_DEF_FAIL_FAST_TMO -1 /* * Define a host as seen by the target. We allocate one at boot, but also @@ -59,6 +61,7 @@ enum { NVMF_OPT_CTRL_LOSS_TMO = 1 << 11, NVMF_OPT_HOST_ID = 1 << 12, NVMF_OPT_DUP_CONNECT = 1 << 13, + NVMF_OPT_FAIL_FAST_TMO = 1 << 20, }; /** @@ -86,6 +89,7 @@ enum { * @max_reconnects: maximum number of allowed reconnect attempts before removing * the controller, (-1) means reconnect forever, zero means remove * immediately; + * @fast_io_fail_tmo: Fast I/O fail timeout in seconds */ struct nvmf_ctrl_options { unsigned mask; @@ -102,6 +106,7 @@ struct nvmf_ctrl_options { unsigned int kato; struct nvmf_host *host; int max_reconnects; + int fast_io_fail_tmo; }; /* diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index f4616df1f7f15..9201946568630 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -198,6 +198,8 @@ static bool nvme_available_path(struct nvme_ns_head *head) struct nvme_ns *ns; list_for_each_entry_rcu(ns, &head->list, siblings) { + if (test_bit(NVME_CTRL_FAILFAST_EXPIRED, &ns->ctrl->flags)) + continue; switch (ns->ctrl->state) { case NVME_CTRL_LIVE: case NVME_CTRL_RESETTING: diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h index 85f814a91ce9c..cd66364eb5429 100644 --- a/drivers/nvme/host/nvme.h +++ b/drivers/nvme/host/nvme.h @@ -211,6 +211,7 @@ struct nvme_ctrl { struct work_struct scan_work; struct work_struct async_event_work; struct delayed_work ka_work; + struct delayed_work failfast_work; struct nvme_command ka_cmd; struct work_struct fw_act_work; unsigned long events; @@ -245,6 +246,8 @@ struct nvme_ctrl { u16 icdoff; u16 maxcmd; int nr_reconnects; + unsigned long flags; +#define NVME_CTRL_FAILFAST_EXPIRED 0 struct nvmf_ctrl_options *opts; struct page *discard_page; -- 2.25.1

1 3