The kdoc for page_pool_get_stats() is missing a Returns: statement.
Looking at this function, I have no idea what is the purpose of
the bool it returns. My guess was that maybe the static inline
stub returns false if CONFIG_PAGE_POOL_STATS=n but such static
inline helper doesn't exist at all. All callers pass a pointer
to a struct on the stack. Make this function void.
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260526155722.2790742-5-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net_mp_open_rxq is currently not used in the tree as all callers are
using __net_mp_open_rxq directly, and net_mp_close_rxq is only used
once while all other locations use __net_mp_close_rxq.
Consolidate into a single API, netif_mp_{open,close}_rxq, using the
netif_ prefix to indicate that the caller is responsible for locking.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Co-developed-by: David Wei <dw@davidwei.uk>
Signed-off-by: David Wei <dw@davidwei.uk>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/20260402231031.447597-6-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The current page_pool alloc-cache size and refill values were chosen to
match the NAPI budget and to leave headroom for XDP_DROP recycling.
These fixed values do not scale well with large pages,
as they significantly increase a given page_pool's memory footprint.
Scale these values to better balance memory footprint across page sizes,
while keeping behavior on 4KB-page systems unchanged.
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Nimrod Oren <noren@nvidia.com>
Link: https://patch.msgid.link/20260309081301.103152-1-noren@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
While testing other changes in vng I noticed that
nl_netdev.page_pool_check flakes. This never happens in real CI.
Turns out vng may boot and get to that test in less than a second.
page_pool_detached() records the detach time in seconds, so if
vng is fast enough detach time is set to 0. Other code treats
0 as "not detached". detach_time is only used to report the state
to the user, so it's not a huge deal in practice but let's fix it.
Store the raw ktime_t (nanoseconds) instead. A nanosecond value
of 0 is practically impossible.
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Fixes: 69cb4952b6 ("net: page_pool: report when page pool was destroyed")
Link: https://patch.msgid.link/20260310003907.3540019-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Allow memory providers to configure rx queues with a custom receive
page size. It's passed in struct pp_memory_provider_params, which is
copied into the queue, so it's preserved across queue restarts. Then,
it's propagated to the driver in a new queue config parameter.
Drivers should explicitly opt into using it by setting
QCFG_RX_PAGE_SIZE, in which case they should implement ndo_default_qcfg,
validate the size on queue restart and honour the current config in case
of a reset.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Page pool can have pages "directly" (locklessly) recycled to it,
if the NAPI that owns the page pool is scheduled to run on the same CPU.
To make this safe we check that the NAPI is disabled while we destroy
the page pool. In most cases NAPI and page pool lifetimes are tied
together so this happens naturally.
The queue API expects the following order of calls:
-> mem_alloc
alloc new pp
-> stop
napi_disable
-> start
napi_enable
-> mem_free
free old pp
Here we allocate the page pool in ->mem_alloc and free in ->mem_free.
But the NAPIs are only stopped between ->stop and ->start. We created
page_pool_disable_direct_recycling() to safely shut down the recycling
in ->stop. This way the page_pool_destroy() call in ->mem_free doesn't
have to worry about recycling any more.
Unfortunately, the page_pool_disable_direct_recycling() is not enough
to deal with failures which necessitate freeing the _new_ page pool.
If we hit a failure in ->mem_alloc or ->stop the new page pool has
to be freed while the NAPI is active (assuming driver attaches the
page pool to an existing NAPI instance and doesn't reallocate NAPIs).
Freeing the new page pool is technically safe because it hasn't been
used for any packets, yet, so there can be no recycling. But the check
in napi_assert_will_not_race() has no way of knowing that. We could
check if page pool is empty but that'd make the check much less likely
to trigger during development.
Add page_pool_enable_direct_recycling(), pairing with
page_pool_disable_direct_recycling(). It will allow us to create the new
page pools in "disabled" state and only enable recycling when we know
the reconfig operation will not fail.
Coincidentally it will also let us re-enable the recycling for the old
pool, if the reconfig failed:
-> mem_alloc (new)
-> stop (old)
# disables direct recycling for old
-> start (new)
# fail!!
-> start (old)
# go back to old pp but direct recycling is lost :(
-> mem_free (new)
The new helper is idempotent to make the life easier for drivers,
which can operate in HDS mode and support zero-copy Rx.
The driver can call the helper twice whether there are two pools
or it has multiple references to a single pool.
Fixes: 40eca00ae6 ("bnxt_en: unlink page pool when stopping Rx queue")
Tested-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250805003654.2944974-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, bnxt_en driver satisfies the requirements of the Device
memory TCP, which is HDS.
So, it implements rx-side Device memory TCP for bnxt_en driver.
It requires only converting the page API to netmem API.
`struct page` of agg rings are changed to `netmem_ref netmem` and
corresponding functions are changed to a variant of netmem API.
It also passes PP_FLAG_ALLOW_UNREADABLE_NETMEM flag to a parameter of
page_pool.
The netmem will be activated only when a user requests devmem TCP.
When netmem is activated, received data is unreadable and netmem is
disabled, received data is readable.
But drivers don't need to handle both cases because netmem core API will
handle it properly.
So, using proper netmem API is enough for drivers.
Device memory TCP can be tested with
tools/testing/selftests/drivers/net/hw/ncdevmem.
This is tested with BCM57504-N425G and firmware version 232.0.155.8/pkg
232.1.132.8.
Reviewed-by: Mina Almasry <almasrymina@google.com>
Tested-by: David Wei <dw@davidwei.uk>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Link: https://patch.msgid.link/20250415052458.1260575-1-ap420073@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When enabling DMA mapping in page_pool, pages are kept DMA mapped until
they are released from the pool, to avoid the overhead of re-mapping the
pages every time they are used. This causes resource leaks and/or
crashes when there are pages still outstanding while the device is torn
down, because page_pool will attempt an unmap through a non-existent DMA
device on the subsequent page return.
To fix this, implement a simple tracking of outstanding DMA-mapped pages
in page pool using an xarray. This was first suggested by Mina[0], and
turns out to be fairly straight forward: We simply store pointers to
pages directly in the xarray with xa_alloc() when they are first DMA
mapped, and remove them from the array on unmap. Then, when a page pool
is torn down, it can simply walk the xarray and unmap all pages still
present there before returning, which also allows us to get rid of the
get/put_device() calls in page_pool. Using xa_cmpxchg(), no additional
synchronisation is needed, as a page will only ever be unmapped once.
To avoid having to walk the entire xarray on unmap to find the page
reference, we stash the ID assigned by xa_alloc() into the page
structure itself, using the upper bits of the pp_magic field. This
requires a couple of defines to avoid conflicting with the
POINTER_POISON_DELTA define, but this is all evaluated at compile-time,
so does not affect run-time performance. The bitmap calculations in this
patch gives the following number of bits for different architectures:
- 23 bits on 32-bit architectures
- 21 bits on PPC64 (because of the definition of ILLEGAL_POINTER_VALUE)
- 32 bits on other 64-bit architectures
Stashing a value into the unused bits of pp_magic does have the effect
that it can make the value stored there lie outside the unmappable
range (as governed by the mmap_min_addr sysctl), for architectures that
don't define ILLEGAL_POINTER_VALUE. This means that if one of the
pointers that is aliased to the pp_magic field (such as page->lru.next)
is dereferenced while the page is owned by page_pool, that could lead to
a dereference into userspace, which is a security concern. The risk of
this is mitigated by the fact that (a) we always clear pp_magic before
releasing a page from page_pool, and (b) this would need a
use-after-free bug for struct page, which can have many other risks
since page->lru.next is used as a generic list pointer in multiple
places in the kernel. As such, with this patch we take the position that
this risk is negligible in practice. For more discussion, see[1].
Since all the tracking added in this patch is performed on DMA
map/unmap, no additional code is needed in the fast path, meaning the
performance overhead of this tracking is negligible there. A
micro-benchmark shows that the total overhead of the tracking itself is
about 400 ns (39 cycles(tsc) 395.218 ns; sum for both map and unmap[2]).
Since this cost is only paid on DMA map and unmap, it seems like an
acceptable cost to fix the late unmap issue. Further optimisation can
narrow the cases where this cost is paid (for instance by eliding the
tracking when DMA map/unmap is a no-op).
The extra memory needed to track the pages is neatly encapsulated inside
xarray, which uses the 'struct xa_node' structure to track items. This
structure is 576 bytes long, with slots for 64 items, meaning that a
full node occurs only 9 bytes of overhead per slot it tracks (in
practice, it probably won't be this efficient, but in any case it should
be an acceptable overhead).
[0] https://lore.kernel.org/all/CAHS8izPg7B5DwKfSuzz-iOop_YRbk3Sd6Y4rX7KBG9DcVJcyWg@mail.gmail.com/
[1] https://lore.kernel.org/r/20250320023202.GA25514@openwall.com
[2] https://lore.kernel.org/r/ae07144c-9295-4c9d-a400-153bb689fe9e@huawei.com
Reported-by: Yonglong Liu <liuyonglong@huawei.com>
Closes: https://lore.kernel.org/r/8743264a-9700-4227-a556-5f931c720211@huawei.com
Fixes: ff7d6b27f8 ("page_pool: refurbish version of page_pool code")
Suggested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Jesper Dangaard Brouer <hawk@kernel.org>
Tested-by: Jesper Dangaard Brouer <hawk@kernel.org>
Tested-by: Qiuling Ren <qren@redhat.com>
Tested-by: Yuying Ma <yuma@redhat.com>
Tested-by: Yonglong Liu <liuyonglong@huawei.com>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20250409-page-pool-track-dma-v9-2-6a9ef2e0cba8@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
devmem code performs a number of safety checks to avoid having
to reimplement all of them in the drivers. Move those to
__net_mp_open_rxq() and reuse that function for binding to make
sure that io_uring ZC also benefits from them.
While at it rename the queue ID variable to rxq_idx in
__net_mp_open_rxq(), we touch most of the relevant lines.
The XArray insertion is reordered after the netdev_rx_queue_restart()
call, otherwise we'd need to duplicate the queue index check
or risk inserting an invalid pointer. The XArray allocation
failures should be extremely rare.
Reviewed-by: Mina Almasry <almasrymina@google.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Fixes: 6e18ed929d ("net: add helpers for setting a memory provider on an rx queue")
Link: https://patch.msgid.link/20250403013405.2827250-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add helpers for memory providers to interact with page pools.
net_mp_niov_{set,clear}_page_pool() serve to [dis]associate a net_iov
with a page pool. If used, the memory provider is responsible to match
"set" calls with "clear" once a net_iov is not going to be used by a page
pool anymore, changing a page pool, etc.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250204215622.695511-10-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Page ppol tried to cache the NAPI ID in page pool info to avoid
having a dependency on the life cycle of the NAPI instance.
Since commit under Fixes the NAPI ID is not populated until
napi_enable() and there's a good chance that page pool is
created before NAPI gets enabled.
Protect the NAPI pointer with the existing page pool mutex,
the reading path already holds it. napi_id itself we need
to READ_ONCE(), it's protected by netdev_lock() which are
not holding in page pool.
Before this patch napi IDs were missing for mlx5:
# ./cli.py --spec netlink/specs/netdev.yaml --dump page-pool-get
[{'id': 144, 'ifindex': 2, 'inflight': 3072, 'inflight-mem': 12582912},
{'id': 143, 'ifindex': 2, 'inflight': 5568, 'inflight-mem': 22806528},
{'id': 142, 'ifindex': 2, 'inflight': 5120, 'inflight-mem': 20971520},
{'id': 141, 'ifindex': 2, 'inflight': 4992, 'inflight-mem': 20447232},
...
After:
[{'id': 144, 'ifindex': 2, 'inflight': 3072, 'inflight-mem': 12582912,
'napi-id': 565},
{'id': 143, 'ifindex': 2, 'inflight': 4224, 'inflight-mem': 17301504,
'napi-id': 525},
{'id': 142, 'ifindex': 2, 'inflight': 4288, 'inflight-mem': 17563648,
'napi-id': 524},
...
Fixes: 86e25f40aa ("net: napi: Add napi_config")
Reviewed-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20250123231620.1086401-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dmabuf dma-addresses should not be dma_sync'd for CPU/device. Typically
its the driver responsibility to dma_sync for CPU, but the driver should
not dma_sync for CPU if the netmem is actually coming from a dmabuf
memory provider.
The page_pool already exposes a helper for dma_sync_for_cpu:
page_pool_dma_sync_for_cpu. Upgrade this existing helper to handle
netmem, and have it skip dma_sync if the memory is from a dmabuf memory
provider. Drivers should migrate to using this helper when adding
support for netmem.
Also minimize the impact on the dma syncing performance for pages. Special
case the dma-sync path for pages to not go through the overhead checks
for dma-syncing and conversion to netmem.
Cc: Alexander Lobakin <aleksander.lobakin@intel.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Signed-off-by: Mina Almasry <almasrymina@google.com>
Link: https://patch.msgid.link/20241211212033.1684197-5-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
