cgroup
| cgroup interface | 源文件 | 解释 |
|---|---|---|
| io.latency | blk-iolatency.c | |
| io.max io.low | blk-throttle.c | |
| io.prio.class | blk-ioprio.c | |
| io.stat | blk-cgroup.c | |
| io.weight | blk-iocost.c | |
| io.pressure | kernel/sched/psi.c |
blk-stat.c 到底是做啥的，应该只是基础组建而已
基本使用和观察
cgcreate -g io:A
echo "8:16 wiops=100" > io.max
cgexec -g io:A dd if=/dev/zero of=/dev/sda bs=1M count=1000
结果 kernel 卡到了这里啊:
cat /proc/200624/stack
[<0>] balance_dirty_pages+0x3be/0xd60
[<0>] balance_dirty_pages_ratelimited_flags+0x2c3/0x3b0
[<0>] generic_perform_write+0x152/0x210
[<0>] __generic_file_write_iter+0xd5/0x190
[<0>] blkdev_write_iter+0x112/0x1b0
[<0>] vfs_write+0x241/0x410
[<0>] ksys_write+0x6f/0xf0
[<0>] do_syscall_64+0x3e/0x90
[<0>] entry_SYSCALL_64_after_hwframe+0x72/0xdc
- iomap_write_iter
- balance_dirty_pages_ratelimited_flags
- balance_dirty_pages
- balance_dirty_pages_ratelimited_flags
参考
- http://docs.kernel.org/admin-guide/cgroup-v2.html
- https://facebookmicrosites.github.io/cgroup2/docs/io-controller.html
iolatency
- iolatency_set_min_lat_nsec
throttle : io.max io.low
#0 blk_cgroup_bio_start (bio=bio@entry=0xffff8881374a7800) at block/blk-cgroup.c:1992
#1 0xffffffff816c0aad in submit_bio_noacct (bio=0xffff8881374a7800) at block/blk-core.c:794
#2 0xffffffff81412289 in submit_bh_wbc (opf=<optimized out>, bh=<optimized out>, wbc=<optimized out>) at fs/buffer.c:2702
#3 0xffffffff81412b8a in submit_bh (bh=0xffff88809e886000, opf=0) at fs/buffer.c:2707
#4 __bh_read (bh=bh@entry=0xffff88809e886000, op_flags=op_flags@entry=0, wait=wait@entry=false) at fs/buffer.c:2972
#5 0xffffffff81415c68 in bh_read_nowait (op_flags=0, bh=0xffff88809e886000) at ./include/linux/buffer_head.h:427
#6 __block_write_begin_int (folio=<optimized out>, pos=pos@entry=624168960, len=len@entry=3, get_block=get_block@entry=0xffffffff816b7f10 <blkdev_get_block>, iomap=iomap@entry=0x0 <fixed_percpu_data>) at fs/buffer.c:2021
#7 0xffffffff81415d41 in __block_write_begin (get_block=0xffffffff816b7f10 <blkdev_get_block>, len=3, pos=624168960, page=0xffffea0002715f80) at fs/buffer.c:2041
#8 block_write_begin (mapping=<optimized out>, pos=624168960, len=3, pagep=0xffffc900402efd88, get_block=0xffffffff816b7f10 <blkdev_get_block>) at fs/buffer.c:2102
#9 0xffffffff812de041 in generic_perform_write (iocb=<optimized out>, i=0xffffc900402efe78) at mm/filemap.c:3772
#10 0xffffffff812e317f in __generic_file_write_iter (iocb=iocb@entry=0xffffc900402efea0, from=from@entry=0xffffc900402efe78) at mm/filemap.c:3900
#11 0xffffffff816b8483 in blkdev_write_iter (iocb=0xffffc900402efea0, from=0xffffc900402efe78) at block/fops.c:534
#12 0xffffffff813c60bd in call_write_iter (iter=0x1000 <cpu_debug_store>, kio=0xffff8881374a7800, file=0xffff888023f85f00) at ./include/linux/fs.h:2186
#13 new_sync_write (ppos=0xffffc900402eff08, len=3, buf=0x40200d "str", filp=0xffff888023f85f00) at fs/read_write.c:491
#14 vfs_write (file=file@entry=0xffff888023f85f00, buf=buf@entry=0x40200d "str", count=count@entry=3, pos=pos@entry=0xffffc900402eff08) at fs/read_write.c:584
#15 0xffffffff813c650e in ksys_write (fd=<optimized out>, buf=0x40200d "str", count=3) at fs/read_write.c:637
#16 0xffffffff8217cbbc in do_syscall_x64 (nr=<optimized out>, regs=0xffffc900402eff58) at arch/x86/entry/common.c:50
#17 do_syscall_64 (regs=0xffffc900402eff58, nr=<optimized out>) at arch/x86/entry/common.c:80
#18 0xffffffff822000ae in entry_SYSCALL_64 () at arch/x86/entry/entry_64.S:120
#0 blk_cgroup_bio_start (bio=bio@entry=0xffff88803ef38d00) at block/blk-cgroup.c:2069
#1 0xffffffff8176c45c in submit_bio_noacct_nocheck (bio=0xffff88803ef38d00) at block/blk-core.c:681
#2 0xffffffff8178ffac in blk_throtl_dispatch_work_fn (work=0xffff888007c6a890) at block/blk-throttle.c:1265
#3 0xffffffff8116a75e in process_one_work (worker=worker@entry=0xffff8880136f8780, work=0xffff888007c6a890) at kernel/workqueue.c:2408
#4 0xffffffff8116ad5e in worker_thread (__worker=0xffff8880136f8780) at kernel/workqueue.c:2555
#5 0xffffffff81172e07 in kthread (_create=0xffff88801590fb40) at kernel/kthread.c:379
#6 0xffffffff81002aec in ret_from_fork () at arch/x86/entry/entry_64.S:308
#7 0x0000000000000000 in ?? ()
这两个 backtrace 没什么关系吧
io.stat
#0 blkcg_iostat_update (blkg=blkg@entry=0xffff8880201d3400, cur=cur@entry=0xffffc900402efaa0, last=0xffffe8ff01c60cb8) at block/blk-cgroup.c:847
#1 0xffffffff816dff61 in blkcg_rstat_flush (css=<optimized out>, cpu=<optimized out>) at block/blk-cgroup.c:894
#2 0xffffffff81205f2f in cgroup_rstat_flush_locked (cgrp=cgrp@entry=0xffffffff82d536f0 <cgrp_dfl_root+16>, may_sleep=may_sleep@entry=false) at kernel/cgroup/rstat.c:205
#3 0xffffffff81206073 in cgroup_rstat_flush_irqsafe (cgrp=0xffffffff82d536f0 <cgrp_dfl_root+16>) at kernel/cgroup/rstat.c:254
#4 0xffffffff81399696 in __mem_cgroup_flush_stats () at mm/memcontrol.c:636
#5 0xffffffff813a2da8 in mem_cgroup_flush_stats () at mm/memcontrol.c:644
#6 mem_cgroup_wb_stats (wb=wb@entry=0xffff888138329000, pfilepages=pfilepages@entry=0xffffc900402efc10, pheadroom=pheadroom@entry=0xffffc900402efc18, pdirty=pdirty@entry=0xffffc900402efcb0, pwriteback=pwriteback@entry=0xffffc900402efc20) at mm/memcontrol.c:4583
#7 0xffffffff812ea500 in balance_dirty_pages (wb=wb@entry=0xffff888138329000, pages_dirtied=32, flags=flags@entry=0) at mm/page-writeback.c:1724
#8 0xffffffff812eb2ed in balance_dirty_pages_ratelimited_flags (mapping=mapping@entry=0xffff888140806508, flags=flags@entry=0) at ./arch/x86/include/asm/current.h:41
#9 0xffffffff812eb40b in balance_dirty_pages_ratelimited (mapping=mapping@entry=0xffff888140806508) at mm/page-writeback.c:2064
#10 0xffffffff812de0c9 in generic_perform_write (iocb=<optimized out>, i=0xffffc900402efe78) at mm/filemap.c:3806
#11 0xffffffff812e317f in __generic_file_write_iter (iocb=iocb@entry=0xffffc900402efea0, from=from@entry=0xffffc900402efe78) at mm/filemap.c:3900
#12 0xffffffff816b8483 in blkdev_write_iter (iocb=0xffffc900402efea0, from=0xffffc900402efe78) at block/fops.c:534
#13 0xffffffff813c60bd in call_write_iter (iter=0xffffc900402efaa0, kio=0xffff8880201d3400, file=0xffff888023f85f00) at ./include/linux/fs.h:2186
#14 new_sync_write (ppos=0xffffc900402eff08, len=3, buf=0x40200d "str", filp=0xffff888023f85f00) at fs/read_write.c:491
#15 vfs_write (file=file@entry=0xffff888023f85f00, buf=buf@entry=0x40200d "str", count=count@entry=3, pos=pos@entry=0xffffc900402eff08) at fs/read_write.c:584
#16 0xffffffff813c650e in ksys_write (fd=<optimized out>, buf=0x40200d "str", count=3) at fs/read_write.c:637
#17 0xffffffff8217cbbc in do_syscall_x64 (nr=<optimized out>, regs=0xffffc900402eff58) at arch/x86/entry/common.c:50
#18 do_syscall_64 (regs=0xffffc900402eff58, nr=<optimized out>) at arch/x86/entry/common.c:80
#19 0xffffffff822000ae in entry_SYSCALL_64 () at arch/x86/entry/entry_64.S:120
似乎存在一个通用框架
enum rq_qos_id {
RQ_QOS_WBT,
RQ_QOS_LATENCY,
RQ_QOS_COST,
};
static struct rq_qos_ops wbt_rqos_ops = {
.throttle = wbt_wait,
.issue = wbt_issue,
.track = wbt_track,
.requeue = wbt_requeue,
.done = wbt_done,
.cleanup = wbt_cleanup,
.queue_depth_changed = wbt_queue_depth_changed,
.exit = wbt_exit,
#ifdef CONFIG_BLK_DEBUG_FS
.debugfs_attrs = wbt_debugfs_attrs,
#endif
};
static struct rq_qos_ops blkcg_iolatency_ops = {
.throttle = blkcg_iolatency_throttle,
.done_bio = blkcg_iolatency_done_bio,
.exit = blkcg_iolatency_exit,
};
static struct rq_qos_ops ioc_rqos_ops = {
.throttle = ioc_rqos_throttle,
.merge = ioc_rqos_merge,
.done_bio = ioc_rqos_done_bio,
.done = ioc_rqos_done,
.queue_depth_changed = ioc_rqos_queue_depth_changed,
.exit = ioc_rqos_exit,
};
当 disk 出现错误，为什么最后是在 wbt_wait 上:
选项 CONFIG_BLK_DEV_THROTTLING 是做什么的
- 2000 行?
qos 的注册内容
两个的注册位置是不同的:
#0 rq_qos_add (rqos=<optimized out>, q=<optimized out>) at block/blk-rq-qos.h:99
#1 wbt_init (q=q@entry=0xffff8881008f9ef0) at block/blk-wbt.c:871
#2 0xffffffff816f8eef in wbt_enable_default (q=q@entry=0xffff8881008f9ef0) at block/blk-wbt.c:673
#3 0xffffffff816c3e03 in blk_register_queue (disk=disk@entry=0xffff888103981400) at block/blk-sysfs.c:820
#4 0xffffffff816d9692 in device_add_disk (parent=parent@entry=0xffff8881008e3010, disk=0xffff888103981400, groups=groups@entry=0xffffffff82e1b8d0 <virtblk_attr_groups>) at block/genhd.c:485
#5 0xffffffff81aa3ba1 in virtblk_probe (vdev=0xffff8881008e3000) at drivers/block/virtio_blk.c:1150
#6 0xffffffff8180ae6e in virtio_dev_probe (_d=0xffff8881008e3010) at drivers/virtio/virtio.c:305
#7 0xffffffff81a76484 in call_driver_probe (drv=0xffffffff82e1b760 <virtio_blk>, dev=0xffff8881008e3010) at drivers/base/dd.c:560
#8 really_probe (dev=dev@entry=0xffff8881008e3010, drv=drv@entry=0xffffffff82e1b760 <virtio_blk>) at drivers/base/dd.c:639
#9 0xffffffff81a766bd in __driver_probe_device (drv=drv@entry=0xffffffff82e1b760 <virtio_blk>, dev=dev@entry=0xffff8881008e3010) at drivers/base/dd.c:778
#10 0xffffffff81a76749 in driver_probe_device (drv=drv@entry=0xffffffff82e1b760 <virtio_blk>, dev=dev@entry=0xffff8881008e3010) at drivers/base/dd.c:808
#11 0xffffffff81a769c5 in __driver_attach (data=0xffffffff82e1b760 <virtio_blk>, dev=0xffff8881008e3010) at drivers/base/dd.c:1194
#12 __driver_attach (dev=0xffff8881008e3010, data=0xffffffff82e1b760 <virtio_blk>) at drivers/base/dd.c:1134
#13 0xffffffff81a73fc7 in bus_for_each_dev (bus=<optimized out>, start=start@entry=0x0 <fixed_percpu_data>, data=data@entry=0xffffffff82e1b760 <virtio_blk>, fn=fn@entry=0xffffffff81a76940 <__driver_attach>) at drivers/base/bus.c:301
#14 0xffffffff81a75e79 in driver_attach (drv=drv@entry=0xffffffff82e1b760 <virtio_blk>) at drivers/base/dd.c:1211
#15 0xffffffff81a75810 in bus_add_driver (drv=drv@entry=0xffffffff82e1b760 <virtio_blk>) at drivers/base/bus.c:618
#16 0xffffffff81a77c2e in driver_register (drv=drv@entry=0xffffffff82e1b760 <virtio_blk>) at drivers/base/driver.c:246
#17 0xffffffff8180a58b in register_virtio_driver (driver=driver@entry=0xffffffff82e1b760 <virtio_blk>) at drivers/virtio/virtio.c:357
#18 0xffffffff835eb0fb in virtio_blk_init () at drivers/block/virtio_blk.c:1284
#19 0xffffffff81001943 in do_one_initcall (fn=0xffffffff835eb0aa <virtio_blk_init>) at init/main.c:1306
#20 0xffffffff8359b818 in do_initcall_level (command_line=0xffff888100129b40 "root", level=6) at init/main.c:1379
#21 do_initcalls () at init/main.c:1395
#22 do_basic_setup () at init/main.c:1414
#23 kernel_init_freeable () at init/main.c:1634
#24 0xffffffff82186485 in kernel_init (unused=<optimized out>) at init/main.c:1522
#25 0xffffffff8100265c in ret_from_fork () at arch/x86/entry/entry_64.S:308
__alloc_disk_node- blkcg_init_disk
- blk_ioprio_init
- blk_throtl_init
- blk_iolatency_init
- blkcg_init_disk
blk-wbt 机制和 memory cgroup 中的 wbt 啥关系啊
什么是 request_list
db6d995235606191fa9db0c717e9d843200b71ea db6d995235606191fa9db0c717e9d843200b71ea
好的，连是什么都不知道，然后就被删除了
The block-throttling low-limit mechanism, described in the Kconfig file as “a best effort limit to prioritize cgroups”, has been removed. It was marked as “experimental” since being introduced in 2017, does not appear to have acquired users, and complicated the maintenance of the block layer.
https://lwn.net/Articles/973687/
本站所有文章转发 CSDN 将按侵权追究法律责任，其它情况随意。