On 16/04/2021 09:39, Ming Lei wrote:
On Fri, Apr 16, 2021 at 09:29:37AM +0100, John Garry wrote:
On 16/04/2021 01:46, Ming Lei wrote:
I can't seem to recreate your same issue. Are you mainline defconfig, or a
special disto config?
The config is rhel8 config.
Can you share that? Has anyone tested against mainline x86 config?
Sure, see the attachment.
Thanks. I assume that this is not seen on mainline x86 defconfig.
Unfortunately it's anything but easy for me to install an x86 kernel ATM.
And I am still seeing this on hisi_sas v2 hw with 5.12-rc7:
[ 214.448368] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
[ 214.454468] rcu:Tasks blocked on level-1 rcu_node (CPUs 0-15):
[ 214.460474] (detected by 40, t=5255 jiffies, g=2229, q=1110)
[ 214.466208] rcu: All QSes seen, last rcu_preempt kthread activity 1
(4294945760-4294945759), jiffies_till_next_fqs=1, root ->qsmask 0x1
[ 214.478466] BUG: scheduling while atomic:
irq/151-hisi_sa/503/0x00000004
[ 214.485162] Modules linked in:
[ 214.488208] CPU: 40 PID: 503 Comm: irq/151-hisi_sa Not tainted 5.11.0
#75
[ 214.494985] Hardware name: Huawei Taishan 2280 /D05, BIOS Hisilicon
D05 IT21 Nemo 2.0 RC0 04/18/2018
[ 214.504105] Call trace:
[ 214.506540] dump_backtrace+0x0/0x1b0
[ 214.510208] show_stack+0x18/0x68
[ 214.513513] dump_stack+0xd8/0x134
[ 214.516907] __schedule_bug+0x60/0x78
[ 214.520560] __schedule+0x620/0x6d8
[ 214.524039] schedule+0x70/0x108
[ 214.527256] irq_thread+0xdc/0x230
[ 214.530648] kthread+0x154/0x158
[ 214.533866] ret_from_fork+0x10/0x30
john@ubuntu:~$
For rw=randread and mq-deadline only, it seems. v5.11 has the same. Not
sure if this is a driver or other issue.
Today I don't have access to other boards with enough disks to get a
decent throughput to test against :(
Thanks,
John
