On 8/21/20 11:30 AM, Christoph Hellwig wrote:
On Fri, Aug 21, 2020 at 10:15:33AM +0530, Ritesh Harjani wrote:
Please correct me here, but as I see, bio has only these two limits
which it checks for adding page to bio. It doesn't check for limits
of /sys/block/<dev>/queue/* no? I guess then it could be checked
by block layer below b4 submitting the bio?
The bio does not, but the blk-mq code will split the bios when mapping
it to requests, take a look at blk_mq_submit_bio and __blk_queue_split.
Thanks :)
But while the default limits are quite low, they can be increased
siginificantly, which tends to help with performance and is often
also done by scripts shipped by the distributions.
This issue was first observed while running a fio run on a system with
huge memory. But then here is an easy way we figured out to trigger the
issue almost everytime with loop device on my VM setup. I have provided
all the details on this below.
Can you wire this up for xfstests?
Sure, will do that.
I do see generic/460 does play with such vm dirty_ratio params which
this test would also require to manipulate to trigger this issue.
Thanks for the suggestion!
-ritesh
