16fbf79...
by
Linus Torvalds <email address hidden>
Linux 5.6-rc7
67d584e...
by
Linus Torvalds <email address hidden>
Merge tag 'for-5.6-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs fixes from David Sterba:
"Two fixes.
The first is a regression: when dropping some incompat bits the
conditions were reversed. The other is a fix for rename whiteout
potentially leaving stack memory linked to a list"
* tag 'for-5.6-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: fix removal of raid[56|1c34} incompat flags after removing block group
btrfs: fix log context list corruption after rename whiteout error
b3c03db...
by
Linus Torvalds <email address hidden>
Merge branch 'akpm' (patches from Andrew)
Merge misc fixes from Andrew Morton:
"10 fixes"
* emailed patches from Andrew Morton <email address hidden>:
x86/mm: split vmalloc_sync_all()
mm, slub: prevent kmalloc_node crashes and memory leaks
mm/mmu_notifier: silence PROVE_RCU_LIST warnings
epoll: fix possible lost wakeup on epoll_ctl() path
mm: do not allow MADV_PAGEOUT for CoW pages
mm, memcg: throttle allocators based on ancestral memory.high
mm, memcg: fix corruption on 64-bit divisor in memory.high throttling
page-flags: fix a crash at SetPageError(THP_SWAP)
mm/hotplug: fix hot remove failure in SPARSEMEM|!VMEMMAP case
memcg: fix NULL pointer dereference in __mem_cgroup_usage_unregister_event
Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in
__purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in
the vunmap() code-path. While this change was necessary to maintain
correctness on x86-32-pae kernels, it also adds additional cycles for
architectures that don't need it.
Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported
severe performance regressions in micro-benchmarks because it now also
calls the x86-64 implementation of vmalloc_sync_all() on vunmap(). But
the vmalloc_sync_all() implementation on x86-64 is only needed for newly
created mappings.
To avoid the unnecessary work on x86-64 and to gain the performance
back, split up vmalloc_sync_all() into two functions:
* vmalloc_sync_mappings(), and
* vmalloc_sync_unmappings()
Most call-sites to vmalloc_sync_all() only care about new mappings being
synchronized. The only exception is the new call-site added in the
above mentioned commit.
Shile Zhang directed us to a report of an 80% regression in reaim
throughput.
This only happens with a mmotm patch "mm/memcontrol.c: allocate
shrinker_map on appropriate NUMA node" [2] which effectively calls
kmalloc_node for each possible node. SLUB however only allocates
kmem_cache_node on online N_NORMAL_MEMORY nodes, and relies on
node_to_mem_node to return such valid node for other nodes since commit
a561ce00b09e ("slub: fall back to node_to_mem_node() node if allocating
on memoryless node"). This is however not true in this configuration
where the _node_numa_mem_ array is not initialized for nodes 0 and 2-31,
thus it contains zeroes and get_partial() ends up accessing
non-allocated kmem_cache_node.
A related issue was reported by Bharata (originally by Ramachandran) [3]
where a similar PowerPC configuration, but with mainline kernel without
patch [2] ends up allocating large amounts of pages by kmalloc-1k
kmalloc-512. This seems to have the same underlying issue with
node_to_mem_node() not behaving as expected, and might probably also
lead to an infinite loop with CONFIG_SLUB_CPU_PARTIAL [4].
This patch should fix both issues by not relying on node_to_mem_node()
anymore and instead simply falling back to NUMA_NO_NODE, when
kmalloc_node(node) is attempted for a node that's not online, or has no
usable memory. The "usable memory" condition is also changed from
node_present_pages() to N_NORMAL_MEMORY node state, as that is exactly
the condition that SLUB uses to allocate kmem_cache_node structures.
The check in get_partial() is removed completely, as the checks in
___slab_alloc() are now sufficient to prevent get_partial() being
reached with an invalid node.
It is safe to traverse mm->notifier_subscriptions->list either under
SRCU read lock or mm->notifier_subscriptions->lock using
hlist_for_each_entry_rcu(). Silence the PROVE_RCU_LIST false positives,
for example,
Signed-off-by: Qian Cai <email address hidden>
Signed-off-by: Andrew Morton <email address hidden>
Reviewed-by: Paul E. McKenney <email address hidden>
Reviewed-by: Jason Gunthorpe <email address hidden>
Link: http://<email address hidden>
Signed-off-by: Linus Torvalds <email address hidden>
1b53734...
by
Roman Penyaev <email address hidden>
epoll: fix possible lost wakeup on epoll_ctl() path
This fixes possible lost wakeup introduced by commit a218cc491420.
Originally modifications to ep->wq were serialized by ep->wq.lock, but
in commit a218cc491420 ("epoll: use rwlock in order to reduce
ep_poll_callback() contention") a new rw lock was introduced in order to
relax fd event path, i.e. callers of ep_poll_callback() function.
After the change ep_modify and ep_insert (both are called on epoll_ctl()
path) were switched to ep->lock, but ep_poll (epoll_wait) was using
ep->wq.lock on wqueue list modification.
The bug doesn't lead to any wqueue list corruptions, because wake up
path and list modifications were serialized by ep->wq.lock internally,
but actual waitqueue_active() check prior wake_up() call can be
reordered with modifications of ep ready list, thus wake up can be lost.
And yes, can be healed by explicit smp_mb():
list_add_tail(&epi->rdlink, &ep->rdllist);
smp_mb();
if (waitqueue_active(&ep->wq))
wake_up(&ep->wp);
But let's make it simple, thus current patch replaces ep->wq.lock with
the ep->lock for wqueue modifications, thus wake up path always observes
activeness of the wqueue correcty.
Fixes: a218cc491420 ("epoll: use rwlock in order to reduce ep_poll_callback() contention")
Reported-by: Max Neunhoeffer <email address hidden>
Signed-off-by: Roman Penyaev <email address hidden>
Signed-off-by: Andrew Morton <email address hidden>
Tested-by: Max Neunhoeffer <email address hidden>
Cc: Jakub Kicinski <email address hidden>
Cc: Christopher Kohlhoff <email address hidden>
Cc: Davidlohr Bueso <email address hidden>
Cc: Jason Baron <email address hidden>
Cc: Jes Sorensen <email address hidden>
Cc: <email address hidden> [5.1+]
Link: http://<email address hidden>
References: https://bugzilla.kernel.org/show_bug.cgi?id=205933
Bisected-by: Max Neunhoeffer <email address hidden>
Signed-off-by: Linus Torvalds <email address hidden>
Jann has brought up a very interesting point [1]. While shared pages
are excluded from MADV_PAGEOUT normally, CoW pages can be easily
reclaimed that way. This can lead to all sorts of hard to debug
problems. E.g. performance problems outlined by Daniel [2].
There are runtime environments where there is a substantial memory
shared among security domains via CoW memory and a easy to reclaim way
of that memory, which MADV_{COLD,PAGEOUT} offers, can lead to either
performance degradation in for the parent process which might be more
privileged or even open side channel attacks.
The feasibility of the latter is not really clear to me TBH but there is
no real reason for exposure at this stage. It seems there is no real
use case to depend on reclaiming CoW memory via madvise at this stage so
it is much easier to simply disallow it and this is what this patch
does. Put it simply MADV_{PAGEOUT,COLD} can operate only on the
exclusively owned memory which is a straightforward semantic.
mm, memcg: throttle allocators based on ancestral memory.high
Prior to this commit, we only directly check the affected cgroup's
memory.high against its usage. However, it's possible that we are being
reclaimed as a result of hitting an ancestor memory.high and should be
penalised based on that, instead.
This patch changes memory.high overage throttling to use the largest
overage in its ancestors when considering how many penalty jiffies to
charge. This makes sure that we penalise poorly behaving cgroups in the
same way regardless of at what level of the hierarchy memory.high was
breached.
Fixes: 0e4b01df8659 ("mm, memcg: throttle allocators when failing reclaim over memory.high")
Reported-by: Johannes Weiner <email address hidden>
Signed-off-by: Chris Down <email address hidden>
Signed-off-by: Andrew Morton <email address hidden>
Acked-by: Johannes Weiner <email address hidden>
Cc: Tejun Heo <email address hidden>
Cc: Michal Hocko <email address hidden>
Cc: Nathan Chancellor <email address hidden>
Cc: Roman Gushchin <email address hidden>
Cc: <email address hidden> [5.4.x+]
Link: http://lkml.kernel.org<email address hidden>
Signed-off-by: Linus Torvalds <email address hidden>