Andrew Morton wrote: > On Wed, 7 Oct 2015 09:30:02 +0200 Michal Hocko <mhocko@xxxxxxxxxx> wrote: > >> On Tue 06-10-15 12:22:25, Andrew Morton wrote: >> > On Tue, 6 Oct 2015 19:01:23 +0200 Michal Hocko <mhocko@xxxxxxxxxx> wrote: >> > >> > > On Mon 05-10-15 14:44:22, Shaohua Li wrote: >> > > > The page_counter_memparse() returns pages for the threshold, while >> > > > mem_cgroup_usage() returns bytes for memory usage. Convert the threshold >> > > > to bytes. >> > > > >> > > > Looks a regression introduced by 3e32cb2e0a12b69150 >> > > >> > > Yes. This suggests >> > > Cc: stable # 3.19+ >> > >> > But it's been this way for 2 years and nobody noticed it. How come? >> >> Maybe we do not have that many users of this API with newer kernels. > > Either it's zero or all the users have worked around this bug. > >> > Or at least, nobody reported it. Maybe people *have* noticed it, and >> > adjusted their userspace appropriately. In which case this patch will >> > cause breakage. >> >> I dunno, I would rather have it fixed than keep bug to bug compatibility >> because they would eventually move to a newer kernel one day when they >> see the "breakage" anyway. > > They'd only see breakage if we fixed this in the newer kernel. > > We could just change the docs and leave it as-is. That it is called > "usage_in_bytes" makes that a bit awkward. > > A bit of googling indicates that people are using usage_in_bytes. A > few. All the discussions I found clearly predate this bug. > > So did people just stop using this? Is there some alternative way of > getting the same info? We (Google) are using byte based notifications on memory.limit_in_bytes on a pre 3e32cb2e0a12b69150 kernel. So we'd notice the regression when running newer kernels. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>