From: Yu Kuai yukuai3@huawei.com
hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I58VE5 CVE: NA
--------------------------------
If new configuration is submitted while a bio is throttled, then new waiting time is recaculated regardless that the bio might aready wait for some time:
tg_conf_updated throtl_start_new_slice tg_update_disptime throtl_schedule_next_dispatch
Then io hung can be triggered by always submmiting new configuration before the throttled bio is dispatched.
Fix the problem by respecting the time that throttled bio aready waited. In order to do that, instead of start new slice in tg_conf_updated(), just update 'bytes_disp' and 'io_disp' based on the new configuration.
Signed-off-by: Yu Kuai yukuai3@huawei.com Reviewed-by: Jason Yan yanaijie@huawei.com Signed-off-by: Zheng Zengkai zhengzengkai@huawei.com --- block/blk-throttle.c | 79 ++++++++++++++++++++++++++++++++++++-------- 1 file changed, 66 insertions(+), 13 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c index 95a13da1f343..0427c9c63e81 100644 --- a/block/blk-throttle.c +++ b/block/blk-throttle.c @@ -1421,7 +1421,57 @@ static int tg_print_conf_uint(struct seq_file *sf, void *v) return 0; }
-static void tg_conf_updated(struct throtl_grp *tg, bool global) +static u64 throtl_update_bytes_disp(u64 dispatched, u64 new_limit, + u64 old_limit) +{ + if (new_limit == old_limit) + return dispatched; + + if (!dispatched) + return 0; + + /* + * In the case that multiply will overflow, just return 0. It will only + * let bios to be dispatched earlier. + */ + if (div64_u64(U64_MAX, dispatched) < new_limit) + return 0; + + dispatched *= new_limit; + return div64_u64(dispatched, old_limit); +} + +static u32 throtl_update_io_disp(u32 dispatched, u32 new_limit, u32 old_limit) +{ + if (new_limit == old_limit) + return dispatched; + + if (!dispatched) + return 0; + /* + * In the case that multiply will overflow, just return 0. It will only + * let bios to be dispatched earlier. + */ + if (UINT_MAX / dispatched < new_limit) + return 0; + + dispatched *= new_limit; + return dispatched / old_limit; +} + +static void throtl_update_slice(struct throtl_grp *tg, u64 *old_limits) +{ + tg->bytes_disp[READ] = throtl_update_bytes_disp(tg->bytes_disp[READ], + tg_bps_limit(tg, READ), old_limits[0]); + tg->bytes_disp[WRITE] = throtl_update_bytes_disp(tg->bytes_disp[WRITE], + tg_bps_limit(tg, WRITE), old_limits[1]); + tg->io_disp[READ] = throtl_update_io_disp(tg->io_disp[READ], + tg_iops_limit(tg, READ), (u32)old_limits[2]); + tg->io_disp[WRITE] = throtl_update_io_disp(tg->io_disp[WRITE], + tg_iops_limit(tg, WRITE), (u32)old_limits[3]); +} + +static void tg_conf_updated(struct throtl_grp *tg, u64 *old_limits, bool global) { struct throtl_service_queue *sq = &tg->service_queue; struct cgroup_subsys_state *pos_css; @@ -1460,16 +1510,7 @@ static void tg_conf_updated(struct throtl_grp *tg, bool global) parent_tg->latency_target); }
- /* - * We're already holding queue_lock and know @tg is valid. Let's - * apply the new config directly. - * - * Restart the slices for both READ and WRITES. It might happen - * that a group's limit are dropped suddenly and we don't want to - * account recently dispatched IO with new low rate. - */ - throtl_start_new_slice(tg, READ); - throtl_start_new_slice(tg, WRITE); + throtl_update_slice(tg, old_limits);
if (tg->flags & THROTL_TG_PENDING) { tg_update_disptime(tg); @@ -1502,6 +1543,14 @@ static inline int throtl_restart_syscall_when_busy(int errno) return ret; }
+static void tg_get_limits(struct throtl_grp *tg, u64 *limits) +{ + limits[0] = tg_bps_limit(tg, READ); + limits[1] = tg_bps_limit(tg, WRITE); + limits[2] = tg_iops_limit(tg, READ); + limits[3] = tg_iops_limit(tg, WRITE); +} + static ssize_t tg_set_conf(struct kernfs_open_file *of, char *buf, size_t nbytes, loff_t off, bool is_u64) { @@ -1510,6 +1559,7 @@ static ssize_t tg_set_conf(struct kernfs_open_file *of, struct throtl_grp *tg; int ret; u64 v; + u64 old_limits[4];
ret = blkg_conf_prep(blkcg, &blkcg_policy_throtl, buf, &ctx); if (ret) @@ -1526,13 +1576,14 @@ static ssize_t tg_set_conf(struct kernfs_open_file *of, v = U64_MAX;
tg = blkg_to_tg(ctx.blkg); + tg_get_limits(tg, old_limits);
if (is_u64) *(u64 *)((void *)tg + of_cft(of)->private) = v; else *(unsigned int *)((void *)tg + of_cft(of)->private) = v;
- tg_conf_updated(tg, false); + tg_conf_updated(tg, old_limits, false); ret = 0; out_finish: blkg_conf_finish(&ctx); @@ -1703,6 +1754,7 @@ static ssize_t tg_set_limit(struct kernfs_open_file *of, struct blkg_conf_ctx ctx; struct throtl_grp *tg; u64 v[4]; + u64 old_limits[4]; unsigned long idle_time; unsigned long latency_time; int ret; @@ -1721,6 +1773,7 @@ static ssize_t tg_set_limit(struct kernfs_open_file *of, v[1] = tg->bps_conf[WRITE][index]; v[2] = tg->iops_conf[READ][index]; v[3] = tg->iops_conf[WRITE][index]; + tg_get_limits(tg, old_limits);
idle_time = tg->idletime_threshold_conf; latency_time = tg->latency_target_conf; @@ -1807,7 +1860,7 @@ static ssize_t tg_set_limit(struct kernfs_open_file *of, tg->td->limit_index = LIMIT_LOW; } else tg->td->limit_index = LIMIT_MAX; - tg_conf_updated(tg, index == LIMIT_LOW && + tg_conf_updated(tg, old_limits, index == LIMIT_LOW && tg->td->limit_valid[LIMIT_LOW]); ret = 0; out_finish: