Currently, percpu memory only exposes allocation and utilization
information via debugfs. This more or less is only really useful for
understanding the fragmentation and allocation information at a per-chunk
level with a few global counters. This is also gated behind a config.
BPF and cgroup, for example, have seen an increase in use causing
increased use of percpu memory. Let's make it easier for someone to
identify how much memory is being used.
This patch adds the "Percpu" stat to meminfo to more easily look up how
much percpu memory is in use. This number includes the cost for all
allocated backing pages and not just insight at the per a unit, per chunk
level. Metadata is excluded. I think excluding metadata is fair because
the backing memory scales with the numbere of cpus and can quickly
outweigh the metadata. It also makes this calculation light.
Link: http://lkml.kernel.org/r/20180807184723.74919-1-dennisszhou@gmail.com
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pul removal of obsolete architecture ports from Arnd Bergmann:
"This removes the entire architecture code for blackfin, cris, frv,
m32r, metag, mn10300, score, and tile, including the associated device
drivers.
I have been working with the (former) maintainers for each one to
ensure that my interpretation was right and the code is definitely
unused in mainline kernels. Many had fond memories of working on the
respective ports to start with and getting them included in upstream,
but also saw no point in keeping the port alive without any users.
In the end, it seems that while the eight architectures are extremely
different, they all suffered the same fate: There was one company in
charge of an SoC line, a CPU microarchitecture and a software
ecosystem, which was more costly than licensing newer off-the-shelf
CPU cores from a third party (typically ARM, MIPS, or RISC-V). It
seems that all the SoC product lines are still around, but have not
used the custom CPU architectures for several years at this point. In
contrast, CPU instruction sets that remain popular and have actively
maintained kernel ports tend to all be used across multiple licensees.
[ See the new nds32 port merged in the previous commit for the next
generation of "one company in charge of an SoC line, a CPU
microarchitecture and a software ecosystem" - Linus ]
The removal came out of a discussion that is now documented at
https://lwn.net/Articles/748074/. Unlike the original plans, I'm not
marking any ports as deprecated but remove them all at once after I
made sure that they are all unused. Some architectures (notably tile,
mn10300, and blackfin) are still being shipped in products with old
kernels, but those products will never be updated to newer kernel
releases.
After this series, we still have a few architectures without mainline
gcc support:
- unicore32 and hexagon both have very outdated gcc releases, but the
maintainers promised to work on providing something newer. At least
in case of hexagon, this will only be llvm, not gcc.
- openrisc, risc-v and nds32 are still in the process of finishing
their support or getting it added to mainline gcc in the first
place. They all have patched gcc-7.3 ports that work to some
degree, but complete upstream support won't happen before gcc-8.1.
Csky posted their first kernel patch set last week, their situation
will be similar
[ Palmer Dabbelt points out that RISC-V support is in mainline gcc
since gcc-7, although gcc-7.3.0 is the recommended minimum - Linus ]"
This really says it all:
2498 files changed, 95 insertions(+), 467668 deletions(-)
* tag 'arch-removal' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic: (74 commits)
MAINTAINERS: UNICORE32: Change email account
staging: iio: remove iio-trig-bfin-timer driver
tty: hvc: remove tile driver
tty: remove bfin_jtag_comm and hvc_bfin_jtag drivers
serial: remove tile uart driver
serial: remove m32r_sio driver
serial: remove blackfin drivers
serial: remove cris/etrax uart drivers
usb: Remove Blackfin references in USB support
usb: isp1362: remove blackfin arch glue
usb: musb: remove blackfin port
usb: host: remove tilegx platform glue
pwm: remove pwm-bfin driver
i2c: remove bfin-twi driver
spi: remove blackfin related host drivers
watchdog: remove bfin_wdt driver
can: remove bfin_can driver
mmc: remove bfin_sdh driver
input: misc: remove blackfin rotary driver
input: keyboard: remove bf54x driver
...
A lot of Kconfig symbols have architecture specific dependencies.
In those cases that depend on architectures we have already removed,
they can be omitted.
Acked-by: Kalle Valo <kvalo@codeaurora.org>
Acked-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
In case of memory deficit and low percpu memory pages,
pcpu_balance_workfn() takes pcpu_alloc_mutex for a long
time (as it makes memory allocations itself and waits
for memory reclaim). If tasks doing pcpu_alloc() are
choosen by OOM killer, they can't exit, because they
are waiting for the mutex.
The patch makes pcpu_alloc() to care about killing signal
and use mutex_lock_killable(), when it's allowed by GFP
flags. This guarantees, a task does not miss SIGKILL
from OOM killer.
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
microblaze build broke due to missing declaration of the
cond_resched() invocation added recently. Let's include linux/sched.h
explicitly.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
When a large BPF percpu map is destroyed, I have seen
pcpu_balance_workfn() holding cpu for hundreds of milliseconds.
On KASAN config and 112 hyperthreads, average time to destroy a chunk
is ~4 ms.
[ 2489.841376] destroy chunk 1 in 4148689 ns
...
[ 2490.093428] destroy chunk 32 in 4072718 ns
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The prior patch added support for passing gfp flags through to the
underlying allocators. This patch allows users to pass along gfp flags
(currently only __GFP_NORETRY and __GFP_NOWARN) to the underlying
allocators. This should allow users to decide if they are ok with
failing allocations recovering in a more graceful way.
Additionally, gfp passing was done as additional flags in the previous
patch. Instead, change this to caller passed semantics. GFP_KERNEL is
also removed as the default flag. It continues to be used for internally
caused underlying percpu allocations.
V2:
Removed gfp_percpu_mask in favor of doing it inline.
Removed GFP_KERNEL as a default flag for __alloc_percpu_gfp.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Percpu memory using the vmalloc area based chunk allocator lazily
populates chunks by first requesting the full virtual address space
required for the chunk and subsequently adding pages as allocations come
through. To ensure atomic allocations can succeed, a workqueue item is
used to maintain a minimum number of empty pages. In certain scenarios,
such as reported in [1], it is possible that physical memory becomes
quite scarce which can result in either a rather long time spent trying
to find free pages or worse, a kernel panic.
This patch adds support for __GFP_NORETRY and __GFP_NOWARN passing them
through to the underlying allocators. This should prevent any
unnecessary panics potentially caused by the workqueue item. The passing
of gfp around is as additional flags rather than a full set of flags.
The next patch will change these to caller passed semantics.
V2:
Added const modifier to gfp flags in the balance path.
Removed an extra whitespace.
[1] https://lkml.org/lkml/2018/2/12/551
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Reported-by: syzbot+adb03f3f0bb57ce3acda@syzkaller.appspotmail.com
Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
At some point the function declaration parameters got out of sync with
the function definitions in percpu-vm.c and percpu-km.c. This patch
makes them match again.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Commit 438a506180 ("percpu: don't forget to free the temporary struct
pcpu_alloc_info") uncovered a problem on the CRIS architecture where
the bootmem allocator is initialized with virtual addresses. Given it
has:
#define __va(x) ((void *)((unsigned long)(x) | 0x80000000))
then things just work out because the end result is the same whether you
give this a physical or a virtual address.
Untill you call memblock_free_early(__pa(address)) that is, because
values from __pa() don't match with the virtual addresses stuffed in the
bootmem allocator anymore.
Avoid freeing the temporary pcpu_alloc_info memory on that architecture
until they fix things up to let the kernel boot like it did before.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: 438a506180 ("percpu: don't forget to free the temporary struct pcpu_alloc_info")
Pull percpu update from Tejun Heo:
"Another minor pull request. It only contains one commit which can
reclaim a bit of memory wasted during boot on UP"
* 'for-4.15' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
percpu: don't forget to free the temporary struct pcpu_alloc_info
Add an option for pcpu_alloc() to support __GFP_NOWARN flag.
Currently, we always throw a warning when size or alignment
is unsupported (and also dump stack on failed allocation
requests). The warning itself is harmless since we return
NULL anyway for any failed request, which callers are
required to handle anyway. However, it becomes harmful when
panic_on_warn is set.
The rationale for the WARN() in pcpu_alloc() is that it can
be tracked when larger than supported allocation requests are
made such that allocations limits can be tweaked if warranted.
This makes sense for in-kernel users, however, there are users
of pcpu allocator where allocation size is derived from user
space requests, e.g. when creating BPF maps. In these cases,
the requests should fail gracefully without throwing a splat.
The current work-around was to check allocation size against
the upper limit of PCPU_MIN_UNIT_SIZE from call-sites for
bailing out prior to a call to pcpu_alloc() in order to
avoid throwing the WARN(). This is bad in multiple ways since
PCPU_MIN_UNIT_SIZE is an implementation detail, and having
the checks on call-sites only complicates the code for no
good reason. Thus, lets fix it generically by supporting the
__GFP_NOWARN flag that users can then use with calling the
__alloc_percpu_gfp() helper instead.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Tejun Heo <tj@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unlike the SMP case, the !SMP case does not free the memory for struct
pcpu_alloc_info allocated in setup_per_cpu_areas(). And to give it a
chance of being reused by the page allocator later, align it to a page
boundary just like its size.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Acked-by: Dennis Zhou <dennisszhou@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The iterator functions pcpu_next_md_free_region and
pcpu_next_fit_region use the block offset to determine if they have
checked the area in the prior iteration. However, this causes an issue
when the block offset is greater than subsequent block contig hints. If
within the iterator it moves to check subsequent blocks, it may fail in
the second predicate due to the block offset not being cleared. Thus,
this causes the allocator to skip over blocks leading to false failures
when allocating from the reserved chunk. While this happens in the
general case as well, it will only fail if it cannot allocate a new
chunk.
This patch resets the block offset to 0 to pass the second predicate
when checking subseqent blocks within the iterator function.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reported-and-tested-by: Luis Henriques <lhenriques@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The other patches contain a lot of information, so adding this
information in a separate patch. It adds my copyright and a brief
explanation of how the bitmap allocator works. There is a minor typo as
well in the prior explanation so that is fixed.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The simple, and expensive, way to find a free area is to iterate over
the entire bitmap until an area is found that fits the allocation size
and alignment. This patch makes use of an iterate that find an area to
check by using the block level contig hints. It will only return an area
that can fit the size and alignment request. If the request can fit
inside a block, it returns the first_free bit to start checking from to
see if it can be fulfilled prior to the contig hint. The pcpu_alloc_area
check has a bound of a block size added in case it is wrong.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The largest free region will either be a block level contig hint or an
aggregate over the left_free and right_free areas of blocks. This is a
much smaller set of free areas that need to be checked than a full
traverse.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The bitmap allocator must keep metadata consistent. The easiest way is
to scan after every allocation for each affected block and the entire
chunk. This is rather expensive.
The free path can take advantage of current contig hints to prevent
scanning within the start and end block. If a scan is needed, it can
be done by scanning backwards from the start and forwards from the end
to identify the entire free area this can be combined with. The blocks
can then be updated by some basic checks rather than complete block
scans.
A chunk scan happens when the freed area makes a page free, a block
free, or spans across blocks. This is necessary as the contig hint at
this point could span across blocks. The check uses the minimum of page
size and the block size to allow for variable sized blocks. There is a
tradeoff here with not updating after every free. It is possible a
contig hint in one block can be merged with the contig hint in the next
block. This means the contig hint can be off by up to a page. However,
if the chunk's contig hint is contained in one block, the contig hint
will be accurate.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Metadata is kept per block to keep track of where the contig hints are.
Scanning can be avoided when the contig hints are not broken. In that
case, left and right contigs have to be managed manually.
This patch changes the allocation path hint updating to only scan when
contig hints are broken.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch makes the contig hint starting offset optimization from the
previous patch as honest as it can be. For both chunk and block starting
offsets, make sure it keeps the starting offset with the best alignment.
The block skip optimization is added in a later patch when the
pcpu_find_block_fit iterator is swapped in.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch adds chunk->contig_bits_start to keep track of the contig
hint's offset and the check to skip the chunk if it does not fit. If
the chunk's contig hint starting offset cannot satisfy an allocation,
the allocator assumes there is enough memory pressure in this chunk to
either use a different chunk or create a new one. This accepts a less
tight packing for a smoother latency curve.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch adds first_bit to keep track of the first free bit in the
bitmap. This hint helps prevent scanning of fully allocated blocks.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch introduces the bitmap metadata blocks and adds the skeleton
of the code that will be used to maintain these blocks. Each chunk's
bitmap is made up of full metadata blocks. These blocks maintain basic
metadata to help prevent scanning unnecssarily to update hints. Full
scanning methods are used for the skeleton and will be replaced in the
coming patches. A number of helper functions are added as well to do
conversion of pages to blocks and manage offsets. Comments will be
updated as the final version of each function is added.
There exists a relationship between PAGE_SIZE, PCPU_BITMAP_BLOCK_SIZE,
the region size, and unit_size. Every chunk's region (including offsets)
is page aligned at the beginning to preserve alignment. The end is
aligned to LCM(PAGE_SIZE, PCPU_BITMAP_BLOCK_SIZE) to ensure that the end
can fit with the populated page map which is by page and every metadata
block is fully accounted for. The unit_size is already page aligned, but
must also be aligned with PCPU_BITMAP_BLOCK_SIZE to ensure full metadata
blocks.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The percpu memory allocator is experiencing scalability issues when
allocating and freeing large numbers of counters as in BPF.
Additionally, there is a corner case where iteration is triggered over
all chunks if the contig_hint is the right size, but wrong alignment.
This patch replaces the area map allocator with a basic bitmap allocator
implementation. Each subsequent patch will introduce new features and
replace full scanning functions with faster non-scanning options when
possible.
Implementation:
This patchset removes the area map allocator in favor of a bitmap
allocator backed by metadata blocks. The primary goal is to provide
consistency in performance and memory footprint with a focus on small
allocations (< 64 bytes). The bitmap removes the heavy memmove from the
freeing critical path and provides a consistent memory footprint. The
metadata blocks provide a bound on the amount of scanning required by
maintaining a set of hints.
In an effort to make freeing fast, the metadata is updated on the free
path if the new free area makes a page free, a block free, or spans
across blocks. This causes the chunk's contig hint to potentially be
smaller than what it could allocate by up to the smaller of a page or a
block. If the chunk's contig hint is contained within a block, a check
occurs and the hint is kept accurate. Metadata is always kept accurate
on allocation, so there will not be a situation where a chunk has a
later contig hint than available.
Evaluation:
I have primarily done testing against a simple workload of allocation of
1 million objects (2^20) of varying size. Deallocation was done by in
order, alternating, and in reverse. These numbers were collected after
rebasing ontop of a80099a152. I present the worst-case numbers here:
Area Map Allocator:
Object Size | Alloc Time (ms) | Free Time (ms)
----------------------------------------------
4B | 310 | 4770
16B | 557 | 1325
64B | 436 | 273
256B | 776 | 131
1024B | 3280 | 122
Bitmap Allocator:
Object Size | Alloc Time (ms) | Free Time (ms)
----------------------------------------------
4B | 490 | 70
16B | 515 | 75
64B | 610 | 80
256B | 950 | 100
1024B | 3520 | 200
This data demonstrates the inability for the area map allocator to
handle less than ideal situations. In the best case of reverse
deallocation, the area map allocator was able to perform within range
of the bitmap allocator. In the worst case situation, freeing took
nearly 5 seconds for 1 million 4-byte objects. The bitmap allocator
dramatically improves the consistency of the free path. The small
allocations performed nearly identical regardless of the freeing
pattern.
While it does add to the allocation latency, the allocation scenario
here is optimal for the area map allocator. The area map allocator runs
into trouble when it is allocating in chunks where the latter half is
full. It is difficult to replicate this, so I present a variant where
the pages are second half filled. Freeing was done sequentially. Below
are the numbers for this scenario:
Area Map Allocator:
Object Size | Alloc Time (ms) | Free Time (ms)
----------------------------------------------
4B | 4118 | 4892
16B | 1651 | 1163
64B | 598 | 285
256B | 771 | 158
1024B | 3034 | 160
Bitmap Allocator:
Object Size | Alloc Time (ms) | Free Time (ms)
----------------------------------------------
4B | 481 | 67
16B | 506 | 69
64B | 636 | 75
256B | 892 | 90
1024B | 3262 | 147
The data shows a parabolic curve of performance for the area map
allocator. This is due to the memmove operation being the dominant cost
with the lower object sizes as more objects are packed in a chunk and at
higher object sizes, the traversal of the chunk slots is the dominating
cost. The bitmap allocator suffers this problem as well. The above data
shows the inability to scale for the allocation path with the area map
allocator and that the bitmap allocator demonstrates consistent
performance in general.
The second problem of additional scanning can result in the area map
allocator completing in 52 minutes when trying to allocate 1 million
4-byte objects with 8-byte alignment. The same workload takes
approximately 16 seconds to complete for the bitmap allocator.
V2:
Fixed a bug in pcpu_alloc_first_chunk end_offset was setting the bitmap
using bytes instead of bits.
Added a comment to pcpu_cnt_pop_pages to explain bitmap_weight.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The area map allocator only used a bitmap for the backing page state.
The new bitmap allocator will use bitmaps to manage the allocation
region in addition to this.
This patch generalizes the bitmap iterators so they can be reused with
the bitmap allocator.
Signed-off-by: Dennis Zhou <dennisszhou@gmail.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
