It releases the lock, executes some function provided by the caller,
then reacquires the lock. This is preparation for the implementation of
condvars, which will sleep after between unlocking and relocking.
We need an explicit `relock` method for primitives like `SpinLock` that
have an irqsave variant: we use the guard state to determine if the lock
was originally acquired with the regular `lock` function or
`lock_irqsave`.
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Signed-off-by: Wedson Almeida Filho <walmeida@microsoft.com>
Link: https://lore.kernel.org/rust-for-linux/20230412121431.41627-1-wedsonaf@gmail.com/
[ Removed the irqsave bits as discussed. ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Relax netdev chain creation to allow for loading the ruleset, then
adding/deleting devices at a later stage. Hardware offload does not
support for this feature yet.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Rename nft_flowtable_hooks_destroy() by nft_hooks_destroy() to prepare
for netdev chain device updates.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
In most cases, table, name and handle is sufficient for userspace to
identify an object that has been deleted. Skipping unneeded fields in
the netlink attributes in the message saves bandwidth (ie. less chances
of hitting ENOBUFS).
Rules are an exception: the existing userspace monitor code relies on
the rule definition. This exception can be removed by implementing a
rule cache in userspace, this is already supported by the tracing
infrastructure.
Regarding flowtables, incremental deletion of devices is possible.
Skipping a full notification allows userspace to differentiate between
flowtable removal and incremental removal of devices.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Flowtable and netdev chains are bound to one or several netdevice,
extend netlink error reporting to specify the the netdevice that
triggers the error.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Remove EnterFunction and LeaveFunction.
These debugging macros seem well past their use-by date. And seem to
have little value these days. Removing them allows some trivial cleanup
of some exit paths for some functions. These are also included in this
patch. There is likely scope for further cleanup of both debugging and
unwind paths. But let's leave that for another day.
Only intended to change debug output, and only when CONFIG_IP_VS_DEBUG
is enabled. Compile tested only.
Signed-off-by: Simon Horman <horms@kernel.org>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Consistently use array_size() to calculate the size of ip_vs_conn_tab
in bytes.
Flagged by Coccinelle:
WARNING: array_size is already used (line 1498) to compute the same size
No functional change intended.
Compile tested only.
Signed-off-by: Simon Horman <horms@kernel.org>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
In ip_vs_sync_conn_v0() copy is made to struct ip_vs_sync_conn_options.
That structure looks like this:
struct ip_vs_sync_conn_options {
struct ip_vs_seq in_seq;
struct ip_vs_seq out_seq;
};
The source of the copy is the in_seq field of struct ip_vs_conn. Whose
type is struct ip_vs_seq. Thus we can see that the source - is not as
wide as the amount of data copied, which is the width of struct
ip_vs_sync_conn_option.
The copy is safe because the next field in is another struct ip_vs_seq.
Make use of struct_group() to annotate this.
Flagged by gcc-13 as:
In file included from ./include/linux/string.h:254,
from ./include/linux/bitmap.h:11,
from ./include/linux/cpumask.h:12,
from ./arch/x86/include/asm/paravirt.h:17,
from ./arch/x86/include/asm/cpuid.h:62,
from ./arch/x86/include/asm/processor.h:19,
from ./arch/x86/include/asm/timex.h:5,
from ./include/linux/timex.h:67,
from ./include/linux/time32.h:13,
from ./include/linux/time.h:60,
from ./include/linux/stat.h:19,
from ./include/linux/module.h:13,
from net/netfilter/ipvs/ip_vs_sync.c:38:
In function 'fortify_memcpy_chk',
inlined from 'ip_vs_sync_conn_v0' at net/netfilter/ipvs/ip_vs_sync.c:606:3:
./include/linux/fortify-string.h:529:25: error: call to '__read_overflow2_field' declared with attribute warning: detected read beyond size of field (2nd parameter); maybe use struct_group()? [-Werror=attribute-warning]
529 | __read_overflow2_field(q_size_field, size);
|
Compile tested only.
Signed-off-by: Simon Horman <horms@kernel.org>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
pass it as argument instead. This reduces size of traceinfo to
16 bytes. Total stack usage:
nf_tables_core.c:252 nft_do_chain 304 static
While its possible to also pass basechain as argument, doing so
increases nft_do_chaininfo function size.
Unlike pktinfo/verdict/rule the basechain info isn't used in
the expression evaluation path. gcc places it on the stack, which
results in extra push/pop when it gets passed to the trace helpers
as argument rather than as part of the traceinfo structure.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Just pass it as argument to nft_trace_notify. Stack is reduced by 8 bytes:
nf_tables_core.c:256 nft_do_chain 312 static
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
pass it as argument. No change in object size.
stack usage decreases by 8 byte:
nf_tables_core.c:254 nft_do_chain 320 static
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
This helper is inlined, so keep it as small as possible.
If the static key is true, there is only a very small chance
that info->trace is false:
1. tracing was enabled at this very moment, the static key was
updated to active right after nft_do_table was called.
2. tracing was disabled at this very moment.
trace->info is already false, the static key is about to
be patched to false soon.
In both cases, no event will be sent because info->trace
is false (checked in noinline slowpath). info->nf_trace is irrelevant.
The nf_trace update is redunant in this case, but this will only
happen for short duration, when static key flips.
text data bss dec hex filename
old: 2980 192 32 3204 c84 nf_tables_core.o
new: 2964 192 32 3188 c74i nf_tables_core.o
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
We only need to validate tables that saw changes in the current
transaction.
The existing code revalidates all tables, but this isn't needed as
cross-table jumps are not allowed (chains have table scope).
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The ->cleanup callback needs to be removed, this doesn't work anymore as
the transaction mutex is already released in the ->abort function.
Just do it after a successful validation pass, this either happens
from commit or abort phases where transaction mutex is held.
Fixes: f102d66b33 ("netfilter: nf_tables: use dedicated mutex to guard transactions")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Now that the rule trailer/end marker and the rcu head reside in the
same structure, we no longer need to save/restore the chain pointer
when performing/returning from a jump.
We can simply let the trace infra walk the evaluated rule until it
hits the end marker and then fetch the chain pointer from there.
When the rule is NULL (policy tracing), then chain and basechain
pointers were already identical, so just use the basechain.
This cuts size of jumpstack in half, from 256 to 128 bytes in 64bit,
scripts/stackusage says:
nf_tables_core.c:251 nft_do_chain 328 static
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Walk the rule headers until the trailer one (last_bit flag set) instead
of stopping at last_rule address.
This avoids the need to store the address when jumping to another chain.
This cuts size of jumpstack array by one third, on 64bit from
384 to 256 bytes. Still, stack usage is still quite large:
scripts/stackusage:
nf_tables_core.c:258 nft_do_chain 496 static
Next patch will also remove chain pointer.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
In order to free the rules in a chain via call_rcu, the rule array used
to stash a rcu_head and space for a pointer at the end of the rule array.
When the current nft_rule_dp blob format got added in
2c865a8a28 ("netfilter: nf_tables: add rule blob layout"), this results
in a double-trailer:
size (unsigned long)
struct nft_rule_dp
struct nft_expr
...
struct nft_rule_dp
struct nft_expr
...
struct nft_rule_dp (is_last=1) // Trailer
The trailer, struct nft_rule_dp (is_last=1), is not accounted for in size,
so it can be located via start_addr + size.
Because the rcu_head is stored after 'start+size' as well this means the
is_last trailer is *aliased* to the rcu_head (struct nft_rules_old).
This is harmless, because at this time the nft_do_chain function never
evaluates/accesses the trailer, it only checks the address boundary:
for (; rule < last_rule; rule = nft_rule_next(rule)) {
...
But this way the last_rule address has to be stashed in the jump
structure to restore it after returning from a chain.
nft_do_chain stack usage has become way too big, so put it on a diet.
Without this patch is impossible to use
for (; !rule->is_last; rule = nft_rule_next(rule)) {
... because on free, the needed update of the rcu_head will clobber the
nft_rule_dp is_last bit.
Furthermore, also stash the chain pointer in the trailer, this allows
to recover the original chain structure from nf_tables_trace infra
without a need to place them in the jump struct.
After this patch it is trivial to diet the jump stack structure,
done in the next two patches.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
KVM/arm64 fixes for 6.3, part #4
- Plug a buffer overflow due to the use of the user-provided register
width for firmware regs. Outright reject accesses where the
user register width does not match the kernel representation.
- Protect non-atomic RMW operations on vCPU flags against preemption,
as an update to the flags by an intervening preemption could be lost.
MIPS's exit sections are discarded at runtime as well.
Fixes link error:
`.exit.text' referenced in section `__jump_table' of fs/fuse/inode.o:
defined in discarded section `.exit.text' of fs/fuse/inode.o
Fixes: 99cb0d917f ("arch: fix broken BuildID for arm64 and riscv")
Reported-by: "kernelci.org bot" <bot@kernelci.org>
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Instead of having callers care about the mmap_min_addr logic for the
lowest valid mapping address (and some of them getting it wrong), just
move the logic into vm_unmapped_area() itself. One less thing for various
architecture cases (and generic helpers) to worry about.
We should really try to make much more of this be common code, but baby
steps..
Without this, vm_unmapped_area() could return an address below
mmap_min_addr (because some caller forgot about that). That then causes
the mmap machinery to think it has found a workable address, but then
later security_mmap_addr(addr) is unhappy about it and the mmap() returns
with a nonsensical error (EPERM).
The proper action is to either return ENOMEM (if the virtual address space
is exhausted), or try to find another address (ie do a bottom-up search
for free addresses after the top-down one failed).
See commit 2afc745f3e ("mm: ensure get_unmapped_area() returns higher
address than mmap_min_addr"), which fixed this for one call site (the
generic arch_get_unmapped_area_topdown() fallback) but left other cases
alone.
Link: https://lkml.kernel.org/r/20230418214009.1142926-1-Liam.Howlett@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Liam Howlett <liam.howlett@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Some architectures can have their hugetlb pages down at the lowest PTE
level: their huge_pte_alloc() using pte_alloc_map(), but without any
following pte_unmap(). Since none of these arches uses CONFIG_HIGHPTE,
this is not seen as a problem at present; but would become a problem if
forthcoming changes were to add an rcu_read_lock() into pte_offset_map(),
with the rcu_read_unlock() expected in pte_unmap().
Similarly in their huge_pte_offset(): pte_offset_kernel() is good enough
for that, but it's probably less confusing if we define pte_offset_huge()
along with pte_alloc_huge(). Only define them without CONFIG_HIGHPTE: so
there would be a build error to signal if ever more work is needed.
For ease of development, define these now for 6.4-rc1, ahead of any use:
then architectures can integrate patches using them, independent from mm.
Link: https://lkml.kernel.org/r/ae9e7d98-8a3a-cfd9-4762-bcddffdf96cf@google.com
Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <muchun.song@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
In the case of reverse allocation, mas->index and mas->last do not point
to the correct allocation range, which will cause users to get incorrect
allocation results, so fix it. If the user does not use it in a specific
way, this bug will not be triggered.
This is a bug, but only VMA uses it now, the way VMA is used now will
not trigger it. There is a possibility that a user will trigger it in
the future.
Also re-check whether the size is still satisfied after the lower bound
was increased, which is a corner case and is incorrect in previous
versions.
Link: https://lkml.kernel.org/r/20230419093625.99201-1-zhangpeng.00@bytedance.com
Fixes: 54a611b605 ("Maple Tree: add new data structure")
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Cc: Liam R. Howlett <Liam.Howlett@Oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
zsmalloc pool can be compacted concurrently by many contexts,
e.g.
cc1 handle_mm_fault()
do_anonymous_page()
__alloc_pages_slowpath()
try_to_free_pages()
do_try_to_free_pages(
lru_gen_shrink_node()
shrink_slab()
do_shrink_slab()
zs_shrinker_scan()
zs_compact()
Pool compaction is currently (basically) single-threaded as
it is performed under pool->lock. Having multiple compaction
threads results in unnecessary contention, as each thread
competes for pool->lock. This, in turn, affects all zsmalloc
operations such as zs_malloc(), zs_map_object(), zs_free(), etc.
Introduce the pool->compaction_in_progress atomic variable,
which ensures that only one compaction context can run at a
time. This reduces overall pool->lock contention in (corner)
cases when many contexts attempt to shrink zspool simultaneously.
Link: https://lkml.kernel.org/r/20230418074639.1903197-1-senozhatsky@chromium.org
Fixes: c0547d0b6a ("zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks")
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Reviewed-by: Yosry Ahmed <yosryahmed@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
This adds three new tests to the selftests for KSM. These tests use the
new prctl API's to enable and disable KSM.
1) add new prctl flags to prctl header file in tools dir
This adds the new prctl flags to the include file prct.h in the
tools directory. This makes sure they are available for testing.
2) add KSM prctl merge test to ksm_tests
This adds the -t option to the ksm_tests program. The -t flag
allows to specify if it should use madvise or prctl ksm merging.
3) add two functions for debugging merge outcome for ksm_tests
This adds two functions to report the metrics in /proc/self/ksm_stat
and /sys/kernel/debug/mm/ksm. The debug output is enabled with the
-d option.
4) add KSM prctl test to ksm_functional_tests
This adds a test to the ksm_functional_test that verifies that the
prctl system call to enable / disable KSM works.
5) add KSM fork test to ksm_functional_test
Add fork test to verify that the MMF_VM_MERGE_ANY flag is inherited
by the child process.
Link: https://lkml.kernel.org/r/20230418051342.1919757-4-shr@devkernel.io
Signed-off-by: Stefan Roesch <shr@devkernel.io>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Bagas Sanjaya <bagasdotme@gmail.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
This adds the general_profit KSM sysfs knob and the process profit metric
knobs to ksm_stat.
1) expose general_profit metric
The documentation mentions a general profit metric, however this
metric is not calculated. In addition the formula depends on the size
of internal structures, which makes it more difficult for an
administrator to make the calculation. Adding the metric for a better
user experience.
2) document general_profit sysfs knob
3) calculate ksm process profit metric
The ksm documentation mentions the process profit metric and how to
calculate it. This adds the calculation of the metric.
4) mm: expose ksm process profit metric in ksm_stat
This exposes the ksm process profit metric in /proc/<pid>/ksm_stat.
The documentation mentions the formula for the ksm process profit
metric, however it does not calculate it. In addition the formula
depends on the size of internal structures. So it makes sense to
expose it.
5) document new procfs ksm knobs
Link: https://lkml.kernel.org/r/20230418051342.1919757-3-shr@devkernel.io
Signed-off-by: Stefan Roesch <shr@devkernel.io>
Reviewed-by: Bagas Sanjaya <bagasdotme@gmail.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm: process/cgroup ksm support", v9.
So far KSM can only be enabled by calling madvise for memory regions. To
be able to use KSM for more workloads, KSM needs to have the ability to be
enabled / disabled at the process / cgroup level.
Use case 1:
The madvise call is not available in the programming language. An
example for this are programs with forked workloads using a garbage
collected language without pointers. In such a language madvise cannot
be made available.
In addition the addresses of objects get moved around as they are
garbage collected. KSM sharing needs to be enabled "from the outside"
for these type of workloads.
Use case 2:
The same interpreter can also be used for workloads where KSM brings
no benefit or even has overhead. We'd like to be able to enable KSM on
a workload by workload basis.
Use case 3:
With the madvise call sharing opportunities are only enabled for the
current process: it is a workload-local decision. A considerable number
of sharing opportunities may exist across multiple workloads or jobs (if
they are part of the same security domain). Only a higler level entity
like a job scheduler or container can know for certain if its running
one or more instances of a job. That job scheduler however doesn't have
the necessary internal workload knowledge to make targeted madvise
calls.
Security concerns:
In previous discussions security concerns have been brought up. The
problem is that an individual workload does not have the knowledge about
what else is running on a machine. Therefore it has to be very
conservative in what memory areas can be shared or not. However, if the
system is dedicated to running multiple jobs within the same security
domain, its the job scheduler that has the knowledge that sharing can be
safely enabled and is even desirable.
Performance:
Experiments with using UKSM have shown a capacity increase of around 20%.
Here are the metrics from an instagram workload (taken from a machine
with 64GB main memory):
full_scans: 445
general_profit: 20158298048
max_page_sharing: 256
merge_across_nodes: 1
pages_shared: 129547
pages_sharing: 5119146
pages_to_scan: 4000
pages_unshared: 1760924
pages_volatile: 10761341
run: 1
sleep_millisecs: 20
stable_node_chains: 167
stable_node_chains_prune_millisecs: 2000
stable_node_dups: 2751
use_zero_pages: 0
zero_pages_sharing: 0
After the service is running for 30 minutes to an hour, 4 to 5 million
shared pages are common for this workload when using KSM.
Detailed changes:
1. New options for prctl system command
This patch series adds two new options to the prctl system call.
The first one allows to enable KSM at the process level and the second
one to query the setting.
The setting will be inherited by child processes.
With the above setting, KSM can be enabled for the seed process of a cgroup
and all processes in the cgroup will inherit the setting.
2. Changes to KSM processing
When KSM is enabled at the process level, the KSM code will iterate
over all the VMA's and enable KSM for the eligible VMA's.
When forking a process that has KSM enabled, the setting will be
inherited by the new child process.
3. Add general_profit metric
The general_profit metric of KSM is specified in the documentation,
but not calculated. This adds the general profit metric to
/sys/kernel/debug/mm/ksm.
4. Add more metrics to ksm_stat
This adds the process profit metric to /proc/<pid>/ksm_stat.
5. Add more tests to ksm_tests and ksm_functional_tests
This adds an option to specify the merge type to the ksm_tests.
This allows to test madvise and prctl KSM.
It also adds a two new tests to ksm_functional_tests: one to test
the new prctl options and the other one is a fork test to verify that
the KSM process setting is inherited by client processes.
This patch (of 3):
So far KSM can only be enabled by calling madvise for memory regions. To
be able to use KSM for more workloads, KSM needs to have the ability to be
enabled / disabled at the process / cgroup level.
1. New options for prctl system command
This patch series adds two new options to the prctl system call.
The first one allows to enable KSM at the process level and the second
one to query the setting.
The setting will be inherited by child processes.
With the above setting, KSM can be enabled for the seed process of a
cgroup and all processes in the cgroup will inherit the setting.
2. Changes to KSM processing
When KSM is enabled at the process level, the KSM code will iterate
over all the VMA's and enable KSM for the eligible VMA's.
When forking a process that has KSM enabled, the setting will be
inherited by the new child process.
1) Introduce new MMF_VM_MERGE_ANY flag
This introduces the new flag MMF_VM_MERGE_ANY flag. When this flag
is set, kernel samepage merging (ksm) gets enabled for all vma's of a
process.
2) Setting VM_MERGEABLE on VMA creation
When a VMA is created, if the MMF_VM_MERGE_ANY flag is set, the
VM_MERGEABLE flag will be set for this VMA.
3) support disabling of ksm for a process
This adds the ability to disable ksm for a process if ksm has been
enabled for the process with prctl.
4) add new prctl option to get and set ksm for a process
This adds two new options to the prctl system call
- enable ksm for all vmas of a process (if the vmas support it).
- query if ksm has been enabled for a process.
3. Disabling MMF_VM_MERGE_ANY for storage keys in s390
In the s390 architecture when storage keys are used, the
MMF_VM_MERGE_ANY will be disabled.
Link: https://lkml.kernel.org/r/20230418051342.1919757-1-shr@devkernel.io
Link: https://lkml.kernel.org/r/20230418051342.1919757-2-shr@devkernel.io
Signed-off-by: Stefan Roesch <shr@devkernel.io>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The permissions for the files here are swapped as "count" is read-only and
"scan" is write-only. While this doesn't really matter as these
permissions don't stop the files being opened for reading/writing as
appropriate, they are shown by "ls -l" and are confusing.
Link: https://lkml.kernel.org/r/20230418101906.3131303-1-john@metanate.com
Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Staring at the comment "Recheck VMA as permissions can change since
migration started" in remove_migration_pte() can result in confusion,
because if the source PTE/PMD indicates write permissions, then there
should be no need to check VMA write permissions when restoring migration
entries or PTE-mapping a PMD.
Commit d3cb8bf608 ("mm: migrate: Close race between migration completion
and mprotect") introduced the maybe_mkwrite() handling in
remove_migration_pte() in 2014, stating that a race between mprotect() and
migration finishing would be possible, and that we could end up with a
writable PTE that should be readable.
However, mprotect() code first updates vma->vm_flags / vma->vm_page_prot
and then walks the page tables to (a) set all present writable PTEs to
read-only and (b) convert all writable migration entries to readable
migration entries. While walking the page tables and modifying the
entries, migration code has to grab the PT locks to synchronize against
concurrent page table modifications.
Assuming migration would find a writable migration entry (while holding
the PT lock) and replace it with a writable present PTE, surely mprotect()
code didn't stumble over the writable migration entry yet (converting it
into a readable migration entry) and would instead wait for the PT lock to
convert the now present writable PTE into a read-only PTE. As mprotect()
didn't finish yet, the behavior is just like migration didn't happen: a
writable PTE will be converted to a read-only PTE.
So it's fine to rely on the writability information in the source PTE/PMD
and not recheck against the VMA as long as we're holding the PT lock to
synchronize with anyone who concurrently wants to downgrade write
permissions (like mprotect()) by first adjusting vma->vm_flags /
vma->vm_page_prot to then walk over the page tables to adjust the page
table entries.
Running test cases that should reveal such races -- mprotect(PROT_READ)
racing with page migration or THP splitting -- for multiple hours did not
reveal an issue with this cleanup.
Link: https://lkml.kernel.org/r/20230418142113.439494-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
In commit fd4a7ac329 ("mm: migrate: try again if THP split is failed due
to page refcnt"), if the THP splitting fails due to page reference count,
we will retry to improve migration successful rate. But the failed
splitting is counted as migration failure and migration retry, which will
cause duplicated failure counting. So, in this patch, this is fixed via
undoing the failure counting if we decide to retry. The patch is tested
via failure injection.
Link: https://lkml.kernel.org/r/20230416235929.1040194-1-ying.huang@intel.com
Fixes: fd4a7ac329 ("mm: migrate: try again if THP split is failed due to page refcnt")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
