We'll soon (next change in the series) need a fuller unwind path
in page_pool_create() so create the inverse of page_pool_init().
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
To fully benefit from previous commit add one byte of state
in the first cache line recording if we need to look at
the slow part.
The packing isn't all that impressive right now, we create
a 7B hole. I'm expecting Olek's rework will reshuffle this,
anyway.
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Link: https://lore.kernel.org/r/20231121000048.789613-3-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
struct page_pool is rather performance critical and we use
16B of the first cache line to store 2 pointers used only
by test code. Future patches will add more informational
(non-fast path) attributes.
It's convenient for the user of the API to not have to worry
which fields are fast and which are slow path. Use struct
groups to split the params into the two categories internally.
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/r/20231121000048.789613-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When ptr_ring_init() returns failure in page_pool_init(), free_percpu()
is not called to free pool->recycle_stats, which may cause memory
leak.
Fixes: ad6fa1e1ab ("page_pool: Add recycle stats")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Reviewed-by: Yunsheng Lin <linyunsheng@huawei.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/r/20231030091256.2915394-1-shaojijie@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Currently page pool supports the below use cases:
use case 1: allocate page without page splitting using
page_pool_alloc_pages() API if the driver knows
that the memory it need is always bigger than
half of the page allocated from page pool.
use case 2: allocate page frag with page splitting using
page_pool_alloc_frag() API if the driver knows
that the memory it need is always smaller than
or equal to the half of the page allocated from
page pool.
There is emerging use case [1] & [2] that is a mix of the
above two case: the driver doesn't know the size of memory it
need beforehand, so the driver may use something like below to
allocate memory with least memory utilization and performance
penalty:
if (size << 1 > max_size)
page = page_pool_alloc_pages();
else
page = page_pool_alloc_frag();
To avoid the driver doing something like above, add the
page_pool_alloc() API to support the above use case, and update
the true size of memory that is acctually allocated by updating
'*size' back to the driver in order to avoid exacerbating
truesize underestimate problem.
Rename page_pool_free() which is used in the destroy process to
__page_pool_destroy() to avoid confusion with the newly added
API.
1. https://lore.kernel.org/all/d3ae6bd3537fbce379382ac6a42f67e22f27ece2.1683896626.git.lorenzo@kernel.org/
2. https://lore.kernel.org/all/20230526054621.18371-3-liangchen.linux@gmail.com/
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
CC: Lorenzo Bianconi <lorenzo@kernel.org>
CC: Alexander Duyck <alexander.duyck@gmail.com>
CC: Liang Chen <liangchen.linux@gmail.com>
CC: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20231020095952.11055-4-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
PP_FLAG_PAGE_FRAG is not really needed after pp_frag_count
handling is unified and page_pool_alloc_frag() is supported
in 32-bit arch with 64-bit DMA, so remove it.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
CC: Lorenzo Bianconi <lorenzo@kernel.org>
CC: Alexander Duyck <alexander.duyck@gmail.com>
CC: Liang Chen <liangchen.linux@gmail.com>
CC: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20231020095952.11055-3-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently when page_pool_create() is called with
PP_FLAG_PAGE_FRAG flag, page_pool_alloc_pages() is only
allowed to be called under the below constraints:
1. page_pool_fragment_page() need to be called to setup
page->pp_frag_count immediately.
2. page_pool_defrag_page() often need to be called to drain
the page->pp_frag_count when there is no more user will
be holding on to that page.
Those constraints exist in order to support a page to be
split into multi fragments.
And those constraints have some overhead because of the
cache line dirtying/bouncing and atomic update.
Those constraints are unavoidable for case when we need a
page to be split into more than one fragment, but there is
also case that we want to avoid the above constraints and
their overhead when a page can't be split as it can only
hold a fragment as requested by user, depending on different
use cases:
use case 1: allocate page without page splitting.
use case 2: allocate page with page splitting.
use case 3: allocate page with or without page splitting
depending on the fragment size.
Currently page pool only provide page_pool_alloc_pages() and
page_pool_alloc_frag() API to enable the 1 & 2 separately,
so we can not use a combination of 1 & 2 to enable 3, it is
not possible yet because of the per page_pool flag
PP_FLAG_PAGE_FRAG.
So in order to allow allocating unsplit page without the
overhead of split page while still allow allocating split
page we need to remove the per page_pool flag in
page_pool_is_last_frag(), as best as I can think of, it seems
there are two methods as below:
1. Add per page flag/bit to indicate a page is split or
not, which means we might need to update that flag/bit
everytime the page is recycled, dirtying the cache line
of 'struct page' for use case 1.
2. Unify the page->pp_frag_count handling for both split and
unsplit page by assuming all pages in the page pool is split
into a big fragment initially.
As page pool already supports use case 1 without dirtying the
cache line of 'struct page' whenever a page is recyclable, we
need to support the above use case 3 with minimal overhead,
especially not adding any noticeable overhead for use case 1,
and we are already doing an optimization by not updating
pp_frag_count in page_pool_defrag_page() for the last fragment
user, this patch chooses to unify the pp_frag_count handling
to support the above use case 3.
There is no noticeable performance degradation and some
justification for unifying the frag_count handling with this
patch applied using a micro-benchmark testing in [1].
1. https://lore.kernel.org/all/bf2591f8-7b3c-4480-bb2c-31dc9da1d6ac@huawei.com/
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
CC: Lorenzo Bianconi <lorenzo@kernel.org>
CC: Alexander Duyck <alexander.duyck@gmail.com>
CC: Liang Chen <liangchen.linux@gmail.com>
CC: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20231020095952.11055-2-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently page_pool_alloc_frag() is not supported in 32-bit
arch with 64-bit DMA because of the overlap issue between
pp_frag_count and dma_addr_upper in 'struct page' for those
arches, which seems to be quite common, see [1], which means
driver may need to handle it when using fragment API.
It is assumed that the combination of the above arch with an
address space >16TB does not exist, as all those arches have
64b equivalent, it seems logical to use the 64b version for a
system with a large address space. It is also assumed that dma
address is page aligned when we are dma mapping a page aligned
buffer, see [2].
That means we're storing 12 bits of 0 at the lower end for a
dma address, we can reuse those bits for the above arches to
support 32b+12b, which is 16TB of memory.
If we make a wrong assumption, a warning is emitted so that
user can report to us.
1. https://lore.kernel.org/all/20211117075652.58299-1-linyunsheng@huawei.com/
2. https://lore.kernel.org/all/20230818145145.4b357c89@kernel.org/
Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
CC: Lorenzo Bianconi <lorenzo@kernel.org>
CC: Alexander Duyck <alexander.duyck@gmail.com>
CC: Liang Chen <liangchen.linux@gmail.com>
CC: Guillaume Tucker <guillaume.tucker@collabora.com>
CC: Matthew Wilcox <willy@infradead.org>
CC: Linux-MM <linux-mm@kvack.org>
Link: https://lore.kernel.org/r/20231013064827.61135-2-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Page pool use in hardirq is prohibited, add debug checks
to catch misuses. IIRC we previously discussed using
DEBUG_NET_WARN_ON_ONCE() for this, but there were concerns
that people will have DEBUG_NET enabled in perf testing.
I don't think anyone enables lockdep in perf testing,
so use lockdep to avoid pushback and arguing :)
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/20230804180529.2483231-6-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, touching <net/page_pool/types.h> triggers a rebuild of more
than half of the kernel. That's because it's included in
<linux/skbuff.h>. And each new include to page_pool/types.h adds more
[useless] data for the toolchain to process per each source file from
that pile.
In commit 6a5bcd84e8 ("page_pool: Allow drivers to hint on SKB
recycling"), Matteo included it to be able to call a couple of functions
defined there. Then, in commit 57f05bc2ab ("page_pool: keep pp info as
long as page pool owns the page") one of the calls was removed, so only
one was left. It's the call to page_pool_return_skb_page() in
napi_frag_unref(). The function is external and doesn't have any
dependencies. Having very niche page_pool_types.h included only for that
looks like an overkill.
As %PP_SIGNATURE is not local to page_pool.c (was only in the
early submissions), nothing holds this function there. Teleport
page_pool_return_skb_page() to skbuff.c, just next to the main consumer,
skb_pp_recycle(), and rename it to napi_pp_put_page(), as it doesn't
work with skbs at all and the former name tells nothing. The #if guards
here are only to not compile and have it in the vmlinux when not needed
-- both call sites are already guarded.
Now, touching page_pool_types.h only triggers rebuilding of the drivers
using it and a couple of core networking files.
Suggested-by: Jakub Kicinski <kuba@kernel.org> # make skbuff.h less heavy
Suggested-by: Alexander Duyck <alexanderduyck@fb.com> # move to skbuff.c
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/20230804180529.2483231-3-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Split types and pure function declarations from page_pool.h
and add them in page_page/types.h, so that C sources can
include page_pool.h and headers should generally only include
page_pool/types.h as suggested by jakub.
Rename page_pool.h to page_pool/helpers.h to have both in
one place.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/20230804180529.2483231-2-aleksander.lobakin@intel.com
[Jakub: change microsoft/mana, fix kdoc paths in Documentation]
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All struct members of the driver-facing APIs are documented twice,
in the code and under Documentation. This is a bit tedious.
I also get the feeling that a lot of developers will read the header
when coding, rather than the doc. Bring the two a little closer
together by using kdoc for structs and functions.
Using kdoc also gives us links (mentioning a function or struct
in the text gets replaced by a link to its doc).
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Link: https://lore.kernel.org/r/20230802161821.3621985-3-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that page_pool_release_page() is not exported we can
merge it with page_pool_return_page(). I believe that
the "Do not replace this with page_pool_return_page()"
comment was there in case page_pool_return_page() was
not inlined, to avoid two function calls.
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Reviewed-by: Yunsheng Lin <linyunsheng@huawei.com>
Link: https://lore.kernel.org/r/20230720010409.1967072-5-kuba@kernel.org
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There seems to be no user calling page_pool_release_page()
for legit reasons, all the users simply haven't been converted
to skb-based recycling, yet. Previous changes converted them.
Update the docs, and unexport the function.
Link: https://lore.kernel.org/r/20230720010409.1967072-4-kuba@kernel.org
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
page_pool_ring_[un]lock() use in_softirq() to decide which
spin lock variant to use, and when they are called in the
context with in_softirq() being false, spin_lock_bh() is
called in page_pool_ring_lock() while spin_unlock() is
called in page_pool_ring_unlock(), because spin_lock_bh()
has disabled the softirq in page_pool_ring_lock(), which
causes inconsistency for spin lock pair calling.
This patch fixes it by returning in_softirq state from
page_pool_producer_lock(), and use it to decide which
spin lock variant to use in page_pool_producer_unlock().
As pool->ring has both producer and consumer lock, so
rename it to page_pool_producer_[un]lock() to reflect
the actual usage. Also move them to page_pool.c as they
are only used there, and remove the 'inline' as the
compiler may have better idea to do inlining or not.
Fixes: 7886244736 ("net: page_pool: Add bulk support for ptr_ring")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/r/20230522031714.5089-1-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jesper points out that we must prevent recycling into cache
after page_pool_destroy() is called, because page_pool_destroy()
is not synchronized with recycling (some pages may still be
outstanding when destroy() gets called).
I assumed this will not happen because NAPI can't be scheduled
if its page pool is being destroyed. But I missed the fact that
NAPI may get reused. For instance when user changes ring configuration
driver may allocate a new page pool, stop NAPI, swap, start NAPI,
and then destroy the old pool. The NAPI is running so old page
pool will think it can recycle to the cache, but the consumer
at that point is the destroy() path, not NAPI.
To avoid extra synchronization let the drivers do "unlinking"
during the "swap" stage while NAPI is indeed disabled.
Fixes: 8c48eea3ad ("page_pool: allow caching from safely localized NAPI")
Reported-by: Jesper Dangaard Brouer <jbrouer@redhat.com>
Link: https://lore.kernel.org/all/e8df2654-6a5b-3c92-489d-2fe5e444135f@redhat.com/
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/r/20230419182006.719923-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit c519fe9a4f ("bnxt: add dma mapping attributes") added
DMA_ATTR_WEAK_ORDERING to DMA attrs on bnxt. It has since spread
to a few more drivers (possibly as a copy'n'paste).
DMA_ATTR_WEAK_ORDERING only seems to matter on Sparc and PowerPC/cell,
the rarity of these platforms is likely why we never bothered adding
the attribute in the page pool, even though it should be safe to add.
To make the page pool migration in drivers which set this flag less
of a risk (of regressing the precious sparc database workloads or
whatever needed this) let's add DMA_ATTR_WEAK_ORDERING on all
page pool DMA mappings.
We could make this a driver opt-in but frankly I don't think it's
worth complicating the API. I can't think of a reason why device
accesses to packet memory would have to be ordered.
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Acked-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/r/20230417152805.331865-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Recent patches to mlx5 mentioned a regression when moving from
driver local page pool to only using the generic page pool code.
Page pool has two recycling paths (1) direct one, which runs in
safe NAPI context (basically consumer context, so producing
can be lockless); and (2) via a ptr_ring, which takes a spin
lock because the freeing can happen from any CPU; producer
and consumer may run concurrently.
Since the page pool code was added, Eric introduced a revised version
of deferred skb freeing. TCP skbs are now usually returned to the CPU
which allocated them, and freed in softirq context. This places the
freeing (producing of pages back to the pool) enticingly close to
the allocation (consumer).
If we can prove that we're freeing in the same softirq context in which
the consumer NAPI will run - lockless use of the cache is perfectly fine,
no need for the lock.
Let drivers link the page pool to a NAPI instance. If the NAPI instance
is scheduled on the same CPU on which we're freeing - place the pages
in the direct cache.
With that and patched bnxt (XDP enabled to engage the page pool, sigh,
bnxt really needs page pool work :() I see a 2.6% perf boost with
a TCP stream test (app on a different physical core than softirq).
The CPU use of relevant functions decreases as expected:
page_pool_refill_alloc_cache 1.17% -> 0%
_raw_spin_lock 2.41% -> 0.98%
Only consider lockless path to be safe when NAPI is scheduled
- in practice this should cover majority if not all of steady state
workloads. It's usually the NAPI kicking in that causes the skb flush.
The main case we'll miss out on is when application runs on the same
CPU as NAPI. In that case we don't use the deferred skb free path.
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Tested-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We use BH context only for synchronization, so we don't care if it's
actually serving softirq or not.
As a side node, in case of threaded NAPI, in_serving_softirq() will
return false because it's in process context with BH off, making
page_pool_recycle_in_cache() unreachable.
Signed-off-by: Qingfang DENG <qingfang.deng@siflower.com.cn>
Tested-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: David S. Miller <davem@davemloft.net>
Lin, Yang Shi, Anshuman Khandual and Mike Rapoport
- Some kmemleak fixes from Patrick Wang and Waiman Long
- DAMON updates from SeongJae Park
- memcg debug/visibility work from Roman Gushchin
- vmalloc speedup from Uladzislau Rezki
- more folio conversion work from Matthew Wilcox
- enhancements for coherent device memory mapping from Alex Sierra
- addition of shared pages tracking and CoW support for fsdax, from
Shiyang Ruan
- hugetlb optimizations from Mike Kravetz
- Mel Gorman has contributed some pagealloc changes to improve latency
and realtime behaviour.
- mprotect soft-dirty checking has been improved by Peter Xu
- Many other singleton patches all over the place
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQTTMBEPP41GrTpTJgfdBJ7gKXxAjgUCYuravgAKCRDdBJ7gKXxA
jpqSAQDrXSdII+ht9kSHlaCVYjqRFQz/rRvURQrWQV74f6aeiAD+NHHeDPwZn11/
SPktqEUrF1pxnGQxqLh1kUFUhsVZQgE=
=w/UH
-----END PGP SIGNATURE-----
Merge tag 'mm-stable-2022-08-03' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull MM updates from Andrew Morton:
"Most of the MM queue. A few things are still pending.
Liam's maple tree rework didn't make it. This has resulted in a few
other minor patch series being held over for next time.
Multi-gen LRU still isn't merged as we were waiting for mapletree to
stabilize. The current plan is to merge MGLRU into -mm soon and to
later reintroduce mapletree, with a view to hopefully getting both
into 6.1-rc1.
Summary:
- The usual batches of cleanups from Baoquan He, Muchun Song, Miaohe
Lin, Yang Shi, Anshuman Khandual and Mike Rapoport
- Some kmemleak fixes from Patrick Wang and Waiman Long
- DAMON updates from SeongJae Park
- memcg debug/visibility work from Roman Gushchin
- vmalloc speedup from Uladzislau Rezki
- more folio conversion work from Matthew Wilcox
- enhancements for coherent device memory mapping from Alex Sierra
- addition of shared pages tracking and CoW support for fsdax, from
Shiyang Ruan
- hugetlb optimizations from Mike Kravetz
- Mel Gorman has contributed some pagealloc changes to improve
latency and realtime behaviour.
- mprotect soft-dirty checking has been improved by Peter Xu
- Many other singleton patches all over the place"
[ XFS merge from hell as per Darrick Wong in
https://lore.kernel.org/all/YshKnxb4VwXycPO8@magnolia/ ]
* tag 'mm-stable-2022-08-03' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (282 commits)
tools/testing/selftests/vm/hmm-tests.c: fix build
mm: Kconfig: fix typo
mm: memory-failure: convert to pr_fmt()
mm: use is_zone_movable_page() helper
hugetlbfs: fix inaccurate comment in hugetlbfs_statfs()
hugetlbfs: cleanup some comments in inode.c
hugetlbfs: remove unneeded header file
hugetlbfs: remove unneeded hugetlbfs_ops forward declaration
hugetlbfs: use helper macro SZ_1{K,M}
mm: cleanup is_highmem()
mm/hmm: add a test for cross device private faults
selftests: add soft-dirty into run_vmtests.sh
selftests: soft-dirty: add test for mprotect
mm/mprotect: fix soft-dirty check in can_change_pte_writable()
mm: memcontrol: fix potential oom_lock recursion deadlock
mm/gup.c: fix formatting in check_and_migrate_movable_page()
xfs: fail dax mount if reflink is enabled on a partition
mm/memcontrol.c: remove the redundant updating of stats_flush_threshold
userfaultfd: don't fail on unrecognized features
hugetlb_cgroup: fix wrong hugetlb cgroup numa stat
...
Currently NIC packet receiving performance based on page pool deteriorates
occasionally. To analysis the causes of this problem page allocation stats
are collected. Here are the stats when NIC rx performance deteriorates:
bandwidth(Gbits/s) 16.8 6.91
rx_pp_alloc_fast 13794308 21141869
rx_pp_alloc_slow 108625 166481
rx_pp_alloc_slow_h 0 0
rx_pp_alloc_empty 8192 8192
rx_pp_alloc_refill 0 0
rx_pp_alloc_waive 100433 158289
rx_pp_recycle_cached 0 0
rx_pp_recycle_cache_full 0 0
rx_pp_recycle_ring 362400 420281
rx_pp_recycle_ring_full 6064893 9709724
rx_pp_recycle_released_ref 0 0
The rx_pp_alloc_waive count indicates that a large number of pages' numa
node are inconsistent with the NIC device numa node. Therefore these pages
can't be reused by the page pool. As a result, many new pages would be
allocated by __page_pool_alloc_pages_slow which is time consuming. This
causes the NIC rx performance fluctuations.
The main reason of huge numa mismatch pages in page pool is that page pool
uses alloc_pages_bulk_array to allocate original pages. This function is
not suitable for page allocation in NUMA scenario. So this patch uses
alloc_pages_bulk_array_node which has a NUMA id input parameter to ensure
the NUMA consistent between NIC device and allocated pages.
Repeated NIC rx performance tests are performed 40 times. NIC rx bandwidth
is higher and more stable compared to the datas above. Here are three test
stats, the rx_pp_alloc_waive count is zero and rx_pp_alloc_slow which
indicates pages allocated from slow patch is relatively low.
bandwidth(Gbits/s) 93 93.9 93.8
rx_pp_alloc_fast 60066264 61266386 60938254
rx_pp_alloc_slow 16512 16517 16539
rx_pp_alloc_slow_ho 0 0 0
rx_pp_alloc_empty 16512 16517 16539
rx_pp_alloc_refill 473841 481910 481585
rx_pp_alloc_waive 0 0 0
rx_pp_recycle_cached 0 0 0
rx_pp_recycle_cache_full 0 0 0
rx_pp_recycle_ring 29754145 30358243 30194023
rx_pp_recycle_ring_full 0 0 0
rx_pp_recycle_released_ref 0 0 0
Signed-off-by: Jie Wang <wangjie125@huawei.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/r/20220705113515.54342-1-huangguangbin2@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently If use page pool allocation stats to analysis a RX performance
degradation problem. These stats only count for pages allocate from
page_pool_alloc_pages. But nic drivers such as hns3 use
page_pool_dev_alloc_frag to allocate pages, so page stats in this API
should also be counted.
Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce page_pool APIs to report stats through ethtool and reduce
duplicated code in each driver.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds a function page_pool_get_stats which can be used by drivers to obtain
stats for a specified page_pool.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add per-cpu stats tracking page pool recycling events:
- cached: recycling placed page in the page pool cache
- cache_full: page pool cache was full
- ring: page placed into the ptr ring
- ring_full: page released from page pool because the ptr ring was full
- released_refcnt: page released (and not recycled) because refcnt > 1
Signed-off-by: Joe Damato <jdamato@fastly.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add per-pool statistics counters for the allocation path of a page pool.
These stats are incremented in softirq context, so no locking or per-cpu
variables are needed.
This code is disabled by default and a kernel config option is provided for
users who wish to enable them.
The statistics added are:
- fast: successful fast path allocations
- slow: slow path order-0 allocations
- slow_high_order: slow path high order allocations
- empty: ptr ring is empty, so a slow path allocation was forced.
- refill: an allocation which triggered a refill of the cache
- waive: pages obtained from the ptr ring that cannot be added to
the cache due to a NUMA mismatch.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change is meant to permit a driver to perform "fragmenting" of the
page from within the driver instead of the current model which requires
pre-partitioning the page. The main motivation behind this is to support
use cases where the page will be split up by the driver after DMA instead
of before.
With this change it becomes possible to start using page pool to replace
some of the existing use cases where multiple references were being used
for a single page, but the number needed was unknown as the size could be
dynamic.
For example, with this code it would be possible to do something like
the following to handle allocation:
page = page_pool_alloc_pages();
if (!page)
return NULL;
page_pool_fragment_page(page, DRIVER_PAGECNT_BIAS_MAX);
rx_buf->page = page;
rx_buf->pagecnt_bias = DRIVER_PAGECNT_BIAS_MAX;
Then we would process a received buffer by handling it with:
rx_buf->pagecnt_bias--;
Once the page has been fully consumed we could then flush the remaining
instances with:
if (page_pool_defrag_page(page, rx_buf->pagecnt_bias))
continue;
page_pool_put_defragged_page(pool, page -1, !!budget);
The general idea is that we want to have the ability to allocate a page
with excess fragment count and then trim off the unneeded fragments.
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
As page_pool_refill_alloc_cache() is only called by
__page_pool_get_cached(), which assumes non-concurrent access
as suggested by the comment in __page_pool_get_cached(), and
ptr_ring allows concurrent access between consumer and producer,
so remove the spinlock in page_pool_refill_alloc_cache().
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/r/20220107090042.13605-1-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Store the XDP mem ID inside the page_pool struct so it can be retrieved
later for use in bpf_prog_run().
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/bpf/20220103150812.87914-4-toke@redhat.com
Add a new callback function to page_pool that, if set, will be called every
time a new page is allocated. This will be used from bpf_test_run() to
initialise the page data with the data provided by userspace when running
XDP programs with redirect turned on.
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/bpf/20220103150812.87914-3-toke@redhat.com
This reverts commit d00e60ee54.
As reported by Guillaume in [1]:
Enabling LPAE always enables CONFIG_ARCH_DMA_ADDR_T_64BIT
in 32-bit systems, which breaks the bootup proceess when a
ethernet driver is using page pool with PP_FLAG_DMA_MAP flag.
As we were hoping we had no active consumers for such system
when we removed the dma mapping support, and LPAE seems like
a common feature for 32 bits system, so revert it.
1. https://www.spinics.net/lists/netdev/msg779890.html
Fixes: d00e60ee54 ("page_pool: disable dma mapping support for 32-bit arch with 64-bit DMA")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Reported-by: "kernelci.org bot" <bot@kernelci.org>
Tested-by: "kernelci.org bot" <bot@kernelci.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
As the 32-bit arch with 64-bit DMA seems to rare those days,
and page pool might carry a lot of code and complexity for
systems that possibly.
So disable dma mapping support for such systems, if drivers
really want to work on such systems, they have to implement
their own DMA-mapping fallback tracking outside page_pool.
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no need to synchronize the account updating, so
use the relaxed atomic to avoid some memory barrier in the
data path.
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently page pool only support page recycling when there
is only one user of the page, and the split page reusing
implemented in the most driver can not use the page pool as
bing-pong way of reusing requires the multi user support in
page pool.
Those reusing or recycling has below limitations:
1. page from page pool can only be used be one user in order
for the page recycling to happen.
2. Bing-pong way of reusing in most driver does not support
multi desc using different part of the same page in order
to save memory.
So add multi-users support and frag page recycling in page
pool to overcome the above limitation.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For 32 bit systems with 64 bit dma, dma_addr[1] is used to
store the upper 32 bit dma addr, those system should be rare
those days.
For normal system, the dma_addr[1] in 'struct page' is not
used, so we can reuse dma_addr[1] for storing frag count,
which means how many frags this page might be splited to.
In order to simplify the page frag support in the page pool,
the PAGE_POOL_DMA_USE_PP_FRAG_COUNT macro is added to indicate
the 32 bit systems with 64 bit dma, and the page frag support
in page pool is disabled for such system.
The newly added page_pool_set_frag_count() is called to reserve
the maximum frag count before any page frag is passed to the
user. The page_pool_atomic_sub_frag_count_return() is called
when user is done with the page frag.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, page->pp is cleared and set everytime the page
is recycled, which is unnecessary.
So only set the page->pp when the page is added to the page
pool and only clear it when the page is released from the
page pool.
This is also a preparation to support allocating frag page
in page pool.
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
As mentioned in commit c07aea3ef4 ("mm: add a signature in
struct page"):
"The page->signature field is aliased to page->lru.next and
page->compound_head."
And as the comment in page_is_pfmemalloc():
"lru.next has bit 1 set if the page is allocated from the
pfmemalloc reserves. Callers may simply overwrite it if they
do not need to preserve that information."
The page->signature is OR’ed with PP_SIGNATURE when a page is
allocated in page pool, see __page_pool_alloc_pages_slow(),
and page->signature is checked directly with PP_SIGNATURE in
page_pool_return_skb_page(), which might cause resoure leaking
problem for a page from page pool if bit 1 of lru.next is set
for a pfmemalloc page. What happens here is that the original
pp->signature is OR'ed with PP_SIGNATURE after the allocation
in order to preserve any existing bits(such as the bit 1, used
to indicate a pfmemalloc page), so when those bits are present,
those page is not considered to be from page pool and the DMA
mapping of those pages will be left stale.
As bit 0 is for page->compound_head, So mask both bit 0/1 before
the checking in page_pool_return_skb_page(). And we will return
those pfmemalloc pages back to the page allocator after cleaning
up the DMA mapping.
Fixes: 6a5bcd84e8 ("page_pool: Allow drivers to hint on SKB recycling")
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up to now several high speed NICs have custom mechanisms of recycling
the allocated memory they use for their payloads.
Our page_pool API already has recycling capabilities that are always
used when we are running in 'XDP mode'. So let's tweak the API and the
kernel network stack slightly and allow the recycling to happen even
during the standard operation.
The API doesn't take into account 'split page' policies used by those
drivers currently, but can be extended once we have users for that.
The idea is to be able to intercept the packet on skb_release_data().
If it's a buffer coming from our page_pool API recycle it back to the
pool for further usage or just release the packet entirely.
To achieve that we introduce a bit in struct sk_buff (pp_recycle:1) and
a field in struct page (page->pp) to store the page_pool pointer.
Storing the information in page->pp allows us to recycle both SKBs and
their fragments.
We could have skipped the skb bit entirely, since identical information
can bederived from struct page. However, in an effort to affect the free path
as less as possible, reading a single bit in the skb which is already
in cache, is better that trying to derive identical information for the
page stored data.
The driver or page_pool has to take care of the sync operations on it's own
during the buffer recycling since the buffer is, after opting-in to the
recycling, never unmapped.
Since the gain on the drivers depends on the architecture, we are not
enabling recycling by default if the page_pool API is used on a driver.
In order to enable recycling the driver must call skb_mark_for_recycle()
to store the information we need for recycling in page->pp and
enabling the recycling bit, or page_pool_store_mem_info() for a fragment.
Co-developed-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Co-developed-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is needed by the page_pool to avoid recycling a page not allocated
via page_pool.
The page->signature field is aliased to page->lru.next and
page->compound_head, but it can't be set by mistake because the
signature value is a bad pointer, and can't trigger a false positive
in PageTail() because the last bit is 0.
Co-developed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
32-bit architectures which expect 8-byte alignment for 8-byte integers and
need 64-bit DMA addresses (arm, mips, ppc) had their struct page
inadvertently expanded in 2019. When the dma_addr_t was added, it forced
the alignment of the union to 8 bytes, which inserted a 4 byte gap between
'flags' and the union.
Fix this by storing the dma_addr_t in one or two adjacent unsigned longs.
This restores the alignment to that of an unsigned long. We always
store the low bits in the first word to prevent the PageTail bit from
being inadvertently set on a big endian platform. If that happened,
get_user_pages_fast() racing against a page which was freed and
reallocated to the page_pool could dereference a bogus compound_head(),
which would be hard to trace back to this cause.
Link: https://lkml.kernel.org/r/20210510153211.1504886-1-willy@infradead.org
Fixes: c25fff7171 ("mm: add dma_addr_t to struct page")
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Matteo Croce <mcroce@linux.microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are cases where the page_pool need to refill with pages from the
page allocator. Some workloads cause the page_pool to release pages
instead of recycling these pages.
For these workload it can improve performance to bulk alloc pages from the
page-allocator to refill the alloc cache.
For XDP-redirect workload with 100G mlx5 driver (that use page_pool)
redirecting xdp_frame packets into a veth, that does XDP_PASS to create an
SKB from the xdp_frame, which then cannot return the page to the
page_pool.
Performance results under GitHub xdp-project[1]:
[1] https://github.com/xdp-project/xdp-project/blob/master/areas/mem/page_pool06_alloc_pages_bulk.org
Mel: The patch "net: page_pool: convert to use alloc_pages_bulk_array
variant" was squashed with this patch. From the test page, the array
variant was superior with one of the test results as follows.
Kernel XDP stats CPU pps Delta
Baseline XDP-RX CPU total 3,771,046 n/a
List XDP-RX CPU total 3,940,242 +4.49%
Array XDP-RX CPU total 4,249,224 +12.68%
Link: https://lkml.kernel.org/r/20210325114228.27719-10-mgorman@techsingularity.net
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Alexander Lobakin <alobakin@pm.me>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: David Miller <davem@davemloft.net>
Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In preparation for next patch, move the dma mapping into its own function,
as this will make it easier to follow the changes.
[ilias.apalodimas: make page_pool_dma_map return boolean]
Link: https://lkml.kernel.org/r/20210325114228.27719-9-mgorman@techsingularity.net
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Alexander Lobakin <alobakin@pm.me>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: David Miller <davem@davemloft.net>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
pool_page_reusable() is a leftover from pre-NUMA-aware times. For now,
this function is just a redundant wrapper over page_is_pfmemalloc(),
so inline it into its sole call site.
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Introduce the capability to batch page_pool ptr_ring refill since it is
usually run inside the driver NAPI tx completion loop.
Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Co-developed-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/bpf/08dd249c9522c001313f520796faa777c4089e1c.1605267335.git.lorenzo@kernel.org
page pool API can be useful for non-DMA cases like
xen-netfront driver so let's allow to pass zero flags to
page pool flags.
v2: check DMA direction only if PP_FLAG_DMA_MAP is set
Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Functions starting with __ usually indicate those which are exported,
but should not be called directly. Update some of those declared in the
API and make it more readable.
page_pool_unmap_page() and page_pool_release_page() were doing
exactly the same thing calling __page_pool_clean_page(). Let's
rename __page_pool_clean_page() to page_pool_release_page() and
export it in order to show up on perf logs and get rid of
page_pool_unmap_page().
Finally rename __page_pool_put_page() to page_pool_put_page() since we
can now directly call it from drivers and rename the existing
page_pool_put_page() to page_pool_put_full_page() since they do the same
thing but the latter is trying to sync the full DMA area.
This patch also updates netsec, mvneta and stmmac drivers which use
those functions.
Suggested-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
"do {} while" in page_pool_refill_alloc_cache will always
refill page once whether refill is true or false, and whether
alloc.count of pool is less than PP_ALLOC_CACHE_REFILL or not
this is wrong, and will cause overflow of pool->alloc.cache
the caller of __page_pool_get_cached should provide guarantee
that pool->alloc.cache is safe to access, so in_serving_softirq
should be removed as suggested by Jesper Dangaard Brouer in
https://patchwork.ozlabs.org/patch/1233713/
so fix this issue by calling page_pool_refill_alloc_cache()
only when pool->alloc.count is zero
Fixes: 44768decb7 ("page_pool: handle page recycle for NUMA_NO_NODE condition")
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
