On Wed, 09 Aug 2017 16:58:26 -0700, Greg KH wrote:
> On Wed, Aug 09, 2017 at 11:33:30PM +0000, Holger Hoffstätte wrote:
>> On Wed, 09 Aug 2017 11:12:53 -0700, Greg Kroah-Hartman wrote:
>>
>> > This is the start of the stable review cycle for the 4.9.42 release.
>> > There are 93 patches in this series, all will be posted as a response
>> > to this one. If anyone has any issues with these being applied, please
>> > let me know.
>>
>> The patch "workqueue-restore-wq_unbound-max_active-1-to-be-ordered.patch"
>> creates a regression because some workqueue users diddle with
>> workqueue_set_max_active(), which results in endless WARNING spam -
>> in my case during btrfs metadata operations:
>>
>> Aug 10 00:23:50 tux kernel: WARNING: CPU: 3 PID: 29080 at kernel/workqueue.c:4116 workqueue_set_max_active+0x89/0x90
>> Aug 10 00:23:50 tux kernel: Modules linked in: nfsd auth_rpcgss oid_registry lockd grace sunrpc sch_fq_codel btrfs xor raid6_pq xfs x86_pkg_temp_thermal crc32_pclmul crc32c_intel snd_hda_codec_hdmi nct6775 hwmon_vid coretemp aesni_intel aes_x86_64 glue_helper snd_hda_codec_realtek lrw gf128mul ablk_helper snd_hda_codec_generic cryptd usbhid r8169 mii i2c_i801 i2c_smbus i2c_core snd_hda_intel snd_hda_codec snd_hda_core snd_pcm snd_timer snd soundcore video
>> Aug 10 00:23:50 tux kernel: CPU: 3 PID: 29080 Comm: kworker/u8:5 Tainted: G W 4.9.42 #1
>> Aug 10 00:23:50 tux kernel: Hardware name: System manufacturer System Product Name/P8Z68-V LX, BIOS 4105 07/01/2013
>> Aug 10 00:23:50 tux kernel: Workqueue: btrfs-delayed-meta btrfs_delayed_meta_helper [btrfs]
>> Aug 10 00:23:50 tux kernel: ffffc90004dfbd38 ffffffff812fce27 0000000000000000 0000000000000000
>> Aug 10 00:23:50 tux kernel: ffffc90004dfbd78 ffffffff81056eeb 00001014d17b1000 ffff8807f62daf80
>> Aug 10 00:23:50 tux kernel: ffff8807fadf3e00 ffff8807f9d6e100 ffff880599a3a890 0000000000000000
>> Aug 10 00:23:50 tux kernel: Call Trace:
>> Aug 10 00:23:50 tux kernel: [<ffffffff812fce27>] dump_stack+0x4d/0x66
>> Aug 10 00:23:50 tux kernel: [<ffffffff81056eeb>] __warn+0xcb/0xf0
>> Aug 10 00:23:50 tux kernel: [<ffffffff81056fdd>] warn_slowpath_null+0x1d/0x20
>> Aug 10 00:23:50 tux kernel: [<ffffffff8106e359>] workqueue_set_max_active+0x89/0x90
>> Aug 10 00:23:50 tux kernel: [<ffffffffa04b0ba9>] btrfs_worker_helper+0x2e9/0x2f0 [btrfs]
>> Aug 10 00:23:50 tux kernel: [<ffffffffa04b0b32>] ? btrfs_worker_helper+0x272/0x2f0 [btrfs]
>> Aug 10 00:23:50 tux kernel: [<ffffffffa04b0c8e>] btrfs_delayed_meta_helper+0xe/0x10 [btrfs]
>> Aug 10 00:23:50 tux kernel: [<ffffffff8106f5cd>] process_one_work+0x1dd/0x460
>> Aug 10 00:23:50 tux kernel: [<ffffffff810b449b>] ? try_to_del_timer_sync+0x4b/0x60
>> Aug 10 00:23:50 tux kernel: [<ffffffff8106f898>] worker_thread+0x48/0x4d0
>> Aug 10 00:23:50 tux kernel: [<ffffffff8106f850>] ? process_one_work+0x460/0x460
>> Aug 10 00:23:50 tux kernel: [<ffffffff81075497>] kthread+0xd7/0xf0
>> Aug 10 00:23:50 tux kernel: [<ffffffff810753c0>] ? kthread_park+0x60/0x60
>> Aug 10 00:23:50 tux kernel: [<ffffffff815bf092>] ret_from_fork+0x22/0x30
>> Aug 10 00:23:50 tux kernel: ---[ end trace cc35cf747f3aa89e ]-
>>
>> Luckily this can be easily fixed by also applying upstream followup
>> commit 0a94efb5acbb6 ("workqueue: implicit ordered attribute should be
>> overridable"), which is marked approppriately as fix, but apparently
>> wasn't sent to -stable.
>>
>> I applied it on top of .42-rc1 and the WARNING spam is gone, so please
>> add it to the final .42 release as well.
>
> Should it also go into all of the other stable trees that this commit is
> now going into?
I think so, since the followup is marked as explict fix.
A quick grep for callers of workqueue_set_max_active() in 4.9.42 shows
only btrfs and fscache, so the impact is limited anyway.
Thanks!
Holger