On 8/1/24 21:51, Jens Axboe wrote:
On 7/31/24 7:01 PM, Chenliang Li wrote:
On Thu, 1 Aug 2024 00:13:10 +0100, Pavel Begunkov wrote:
On 5/31/24 06:20, Chenliang Li wrote:
Add a test file for hugepage registered buffers, to make sure the
fixed buffer coalescing feature works safe and soundly.
Testcases include read/write with single/multiple/unaligned/non-2MB
hugepage fixed buffers, and also a should-not coalesce case where
buffer is a mixture of different size'd pages.
lgtm, would be even better if you can add another patch
testing adding a small buffer on the left size of a hugepage,
i.e. like mmap_mixutre() but the small buffer is on the other
side.
Sure, will add that.
I'll queue up this v3, but please do send that as an incremental
patch as I agree it's a good addition
I have to add that tests are already well made, the kernel code
is not symmetric however, so it'd give us more confidence and
help to not regress in the future.
--
Pavel Begunkov
