Hi Linus, This pull request contains a fix for sporadically failing atomic percpu allocations. I only caught it recently while I was reviewing a new series [1] and simultaneously saw reports by btrfs in xfstests [2] and [3]. In v5.9, memcg accounting was extended to percpu done by adding a second type of chunk. I missed an interaction with the free page float count used to ensure we can support atomic allocations. If 1 type of chunk has no free pages, but the other has enough to satisfy the free page float requirement, we will not repopulate the free pages for the former type of chunk. This led to sporadically failing atomic allocations. [1] https://lore.kernel.org/linux-mm/20210324190626.564297-1-guro@xxxxxx/ [2] https://lore.kernel.org/linux-mm/20210401185158.3275.409509F4@xxxxxxxxxxxx/ [3] https://lore.kernel.org/linux-mm/CAL3q7H5RNBjCi708GH7jnczAOe0BLnacT9C+OBgA-Dx9jhB6SQ@xxxxxxxxxxxxxx/ Thanks, Dennis The following changes since commit e49d033bddf5b565044e2abe4241353959bc9120: Linux 5.12-rc6 (2021-04-04 14:15:36 -0700) are available in the Git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/dennis/percpu.git for-5.12-fixes for you to fetch changes up to 0760fa3d8f7fceeea508b98899f1c826e10ffe78: percpu: make pcpu_nr_empty_pop_pages per chunk type (2021-04-09 13:58:38 +0000) ---------------------------------------------------------------- Roman Gushchin (1): percpu: make pcpu_nr_empty_pop_pages per chunk type mm/percpu-internal.h | 2 +- mm/percpu-stats.c | 9 +++++++-- mm/percpu.c | 14 +++++++------- 3 files changed, 15 insertions(+), 10 deletions(-)
