Move queue_id, dev, and need_wakeup from the umem to the
buffer pool. This so that we in a later commit can share the umem
between multiple HW queues. There is one buffer pool per dev and
queue id, so these variables should belong to the buffer pool, not
the umem. Need_wakeup is also something that is set on a per napi
level, so there is usually one per device and queue id. So move
this to the buffer pool too.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-6-git-send-email-magnus.karlsson@intel.com
Move the fill and completion rings from the umem to the buffer
pool. This so that we in a later commit can share the umem
between multiple HW queue ids. In this case, we need one fill and
completion ring per queue id. As the buffer pool is per queue id
and napi id this is a natural place for it and one umem
struture can be shared between these buffer pools.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-5-git-send-email-magnus.karlsson@intel.com
Create and free the buffer pool independently from the umem. Move
these operations that are performed on the buffer pool from the
umem create and destroy functions to new create and destroy
functions just for the buffer pool. This so that in later commits
we can instantiate multiple buffer pools per umem when sharing a
umem between HW queues and/or devices. We also erradicate the
back pointer from the umem to the buffer pool as this will not
work when we introduce the possibility to have multiple buffer
pools per umem.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-4-git-send-email-magnus.karlsson@intel.com
Rename the AF_XDP zero-copy driver interface functions to better
reflect what they do after the replacement of umems with buffer
pools in the previous commit. Mostly it is about replacing the
umem name from the function names with xsk_buff and also have
them take the a buffer pool pointer instead of a umem. The
various ring functions have also been renamed in the process so
that they have the same naming convention as the internal
functions in xsk_queue.h. This so that it will be clearer what
they do and also for consistency.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-3-git-send-email-magnus.karlsson@intel.com
Replace the explicit umem reference passed to the driver in AF_XDP
zero-copy mode with the buffer pool instead. This in preparation for
extending the functionality of the zero-copy mode so that umems can be
shared between queues on the same netdev and also between netdevs. In
this commit, only an umem reference has been added to the buffer pool
struct. But later commits will add other entities to it. These are
going to be entities that are different between different queue ids
and netdevs even though the umem is shared between them.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-2-git-send-email-magnus.karlsson@intel.com
Most of the maps do not use max_entries during verification time.
Thus, those map_meta_equal() do not need to enforce max_entries
when it is inserted as an inner map during runtime. The max_entries
check is removed from the default implementation bpf_map_meta_equal().
The prog_array_map and xsk_map are exception. Its map_gen_lookup
uses max_entries to generate inline lookup code. Thus, they will
implement its own map_meta_equal() to enforce max_entries.
Since there are only two cases now, the max_entries check
is not refactored and stays in its own .c file.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200828011813.1970516-1-kafai@fb.com
Some properties of the inner map is used in the verification time.
When an inner map is inserted to an outer map at runtime,
bpf_map_meta_equal() is currently used to ensure those properties
of the inserting inner map stays the same as the verification
time.
In particular, the current bpf_map_meta_equal() checks max_entries which
turns out to be too restrictive for most of the maps which do not use
max_entries during the verification time. It limits the use case that
wants to replace a smaller inner map with a larger inner map. There are
some maps do use max_entries during verification though. For example,
the map_gen_lookup in array_map_ops uses the max_entries to generate
the inline lookup code.
To accommodate differences between maps, the map_meta_equal is added
to bpf_map_ops. Each map-type can decide what to check when its
map is used as an inner map during runtime.
Also, some map types cannot be used as an inner map and they are
currently black listed in bpf_map_meta_alloc() in map_in_map.c.
It is not unusual that the new map types may not aware that such
blacklist exists. This patch enforces an explicit opt-in
and only allows a map to be used as an inner map if it has
implemented the map_meta_equal ops. It is based on the
discussion in [1].
All maps that support inner map has its map_meta_equal points
to bpf_map_meta_equal in this patch. A later patch will
relax the max_entries check for most maps. bpf_types.h
counts 28 map types. This patch adds 23 ".map_meta_equal"
by using coccinelle. -5 for
BPF_MAP_TYPE_PROG_ARRAY
BPF_MAP_TYPE_(PERCPU)_CGROUP_STORAGE
BPF_MAP_TYPE_STRUCT_OPS
BPF_MAP_TYPE_ARRAY_OF_MAPS
BPF_MAP_TYPE_HASH_OF_MAPS
The "if (inner_map->inner_map_meta)" check in bpf_map_meta_alloc()
is moved such that the same error is returned.
[1]: https://lore.kernel.org/bpf/20200522022342.899756-1-kafai@fb.com/
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200828011806.1970400-1-kafai@fb.com
xsk_getsockopt() is copying uninitialized stack memory to userspace when
'extra_stats' is 'false'. Fix it. Doing '= {};' is sufficient since currently
'struct xdp_statistics' is defined as follows:
struct xdp_statistics {
__u64 rx_dropped;
__u64 rx_invalid_descs;
__u64 tx_invalid_descs;
__u64 rx_ring_full;
__u64 rx_fill_ring_empty_descs;
__u64 tx_ring_empty_descs;
};
When being copied to the userspace, 'stats' will not contain any uninitialized
'holes' between struct fields.
Fixes: 8aa5a33578 ("xsk: Add new statistics")
Suggested-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Peilin Ye <yepeilin.cs@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/bpf/20200728053604.404631-1-yepeilin.cs@gmail.com
Rework the remaining setsockopt code to pass a sockptr_t instead of a
plain user pointer. This removes the last remaining set_fs(KERNEL_DS)
outside of architecture specific code.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Stefan Schmidt <stefan@datenfreihafen.org> [ieee802154]
Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Alexei Starovoitov says:
====================
pull-request: bpf-next 2020-07-13
The following pull-request contains BPF updates for your *net-next* tree.
We've added 36 non-merge commits during the last 7 day(s) which contain
a total of 62 files changed, 2242 insertions(+), 468 deletions(-).
The main changes are:
1) Avoid trace_printk warning banner by switching bpf_trace_printk to use
its own tracing event, from Alan.
2) Better libbpf support on older kernels, from Andrii.
3) Additional AF_XDP stats, from Ciara.
4) build time resolution of BTF IDs, from Jiri.
5) BPF_CGROUP_INET_SOCK_RELEASE hook, from Stanislav.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
It can be useful for the user to know the reason behind a dropped packet.
Introduce new counters which track drops on the receive path caused by:
1. rx ring being full
2. fill ring being empty
Also, on the tx path introduce a counter which tracks the number of times
we attempt pull from the tx ring when it is empty.
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200708072835.4427-2-ciara.loftus@intel.com
Use the dma_need_sync helper instead of (not always entirely correctly)
poking into the dma-mapping internals.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200629130359.2690853-5-hch@lst.de
->dev is already assigned at the top of the function, remove the duplicate
one at the end.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200629130359.2690853-4-hch@lst.de
Invert the polarity and better name the flag so that the use case is
properly documented.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200629130359.2690853-3-hch@lst.de
Set map_btf_name and map_btf_id for all map types so that map fields can
be accessed by bpf programs.
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/a825f808f22af52b018dbe82f1c7d29dab5fc978.1592600985.git.rdna@fb.com
Pull networking fixes from David Miller:
1) Fix cfg80211 deadlock, from Johannes Berg.
2) RXRPC fails to send norigications, from David Howells.
3) MPTCP RM_ADDR parsing has an off by one pointer error, fix from
Geliang Tang.
4) Fix crash when using MSG_PEEK with sockmap, from Anny Hu.
5) The ucc_geth driver needs __netdev_watchdog_up exported, from
Valentin Longchamp.
6) Fix hashtable memory leak in dccp, from Wang Hai.
7) Fix how nexthops are marked as FDB nexthops, from David Ahern.
8) Fix mptcp races between shutdown and recvmsg, from Paolo Abeni.
9) Fix crashes in tipc_disc_rcv(), from Tuong Lien.
10) Fix link speed reporting in iavf driver, from Brett Creeley.
11) When a channel is used for XSK and then reused again later for XSK,
we forget to clear out the relevant data structures in mlx5 which
causes all kinds of problems. Fix from Maxim Mikityanskiy.
12) Fix memory leak in genetlink, from Cong Wang.
13) Disallow sockmap attachments to UDP sockets, it simply won't work.
From Lorenz Bauer.
* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (83 commits)
net: ethernet: ti: ale: fix allmulti for nu type ale
net: ethernet: ti: am65-cpsw-nuss: fix ale parameters init
net: atm: Remove the error message according to the atomic context
bpf: Undo internal BPF_PROBE_MEM in BPF insns dump
libbpf: Support pre-initializing .bss global variables
tools/bpftool: Fix skeleton codegen
bpf: Fix memlock accounting for sock_hash
bpf: sockmap: Don't attach programs to UDP sockets
bpf: tcp: Recv() should return 0 when the peer socket is closed
ibmvnic: Flush existing work items before device removal
genetlink: clean up family attributes allocations
net: ipa: header pad field only valid for AP->modem endpoint
net: ipa: program upper nibbles of sequencer type
net: ipa: fix modem LAN RX endpoint id
net: ipa: program metadata mask differently
ionic: add pcie_print_link_status
rxrpc: Fix race between incoming ACK parser and retransmitter
net/mlx5: E-Switch, Fix some error pointer dereferences
net/mlx5: Don't fail driver on failure to create debugfs
net/mlx5e: CT: Fix ipv6 nat header rewrite actions
...
Alexei Starovoitov says:
====================
pull-request: bpf 2020-06-12
The following pull-request contains BPF updates for your *net* tree.
We've added 26 non-merge commits during the last 10 day(s) which contain
a total of 27 files changed, 348 insertions(+), 93 deletions(-).
The main changes are:
1) sock_hash accounting fix, from Andrey.
2) libbpf fix and probe_mem sanitizing, from Andrii.
3) sock_hash fixes, from Jakub.
4) devmap_val fix, from Jesper.
5) load_bytes_relative fix, from YiFei.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Propagate sock_alloc_send_skb error code, not set it to
EAGAIN unconditionally, when fail to allocate skb, which
might cause that user space unnecessary loops.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1591852266-24017-1-git-send-email-lirongqing@baidu.com
64bit division is kind of expensive, and shift should do the job here.
Signed-off-by: Pavel Machek (CIP) <pavel@denx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
xdp_umem.c had overlapping changes between the 64-bit math fix
for the calculation of npgs and the removal of the zerocopy
memory type which got rid of the chunk_size_nohdr member.
The mlx5 Kconfig conflict is a case where we just take the
net-next copy of the Kconfig entry dependency as it takes on
the ESWITCH dependency by one level of indirection which is
what the 'net' conflicting change is trying to ensure.
Signed-off-by: David S. Miller <davem@davemloft.net>
The npgs member of struct xdp_umem is an u32 entity, and stores the
number of pages the UMEM consumes. The calculation of npgs
npgs = size / PAGE_SIZE
can overflow.
To avoid overflow scenarios, the division is now first stored in a
u64, and the result is verified to fit into 32b.
An alternative would be storing the npgs as a u64, however, this
wastes memory and is an unrealisticly large packet area.
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Reported-by: "Minh Bùi Quang" <minhquangbui99@gmail.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/bpf/CACtPs=GGvV-_Yj6rbpzTVnopgi5nhMoCcTkSkYrJHGQHJWFZMQ@mail.gmail.com/
Link: https://lore.kernel.org/bpf/20200525080400.13195-1-bjorn.topel@gmail.com
In order to reduce the number of function calls, the struct
xsk_buff_pool definition is moved to xsk_buff_pool.h. The functions
xp_get_dma(), xp_dma_sync_for_cpu(), xp_dma_sync_for_device(),
xp_validate_desc() and various helper functions are explicitly
inlined.
Further, move xp_get_handle() and xp_release() to xsk.c, to allow for
the compiler to perform inlining.
rfc->v1: Make sure xp_validate_desc() is inlined for Tx perf. (Maxim)
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200520192103.355233-15-bjorn.topel@gmail.com
There are no users of MEM_TYPE_ZERO_COPY. Remove all corresponding
code, including the "handle" member of struct xdp_buff.
rfc->v1: Fixed spelling in commit message. (Björn)
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200520192103.355233-13-bjorn.topel@gmail.com
In order to simplify AF_XDP zero-copy enablement for NIC driver
developers, a new AF_XDP buffer allocation API is added. The
implementation is based on a single core (single producer/consumer)
buffer pool for the AF_XDP UMEM.
A buffer is allocated using the xsk_buff_alloc() function, and
returned using xsk_buff_free(). If a buffer is disassociated with the
pool, e.g. when a buffer is passed to an AF_XDP socket, a buffer is
said to be released. Currently, the release function is only used by
the AF_XDP internals and not visible to the driver.
Drivers using this API should register the XDP memory model with the
new MEM_TYPE_XSK_BUFF_POOL type.
The API is defined in net/xdp_sock_drv.h.
The buffer type is struct xdp_buff, and follows the lifetime of
regular xdp_buffs, i.e. the lifetime of an xdp_buff is restricted to
a NAPI context. In other words, the API is not replacing xdp_frames.
In addition to introducing the API and implementations, the AF_XDP
core is migrated to use the new APIs.
rfc->v1: Fixed build errors/warnings for m68k and riscv. (kbuild test
robot)
Added headroom/chunk size getter. (Maxim/Björn)
v1->v2: Swapped SoBs. (Maxim)
v2->v3: Initialize struct xdp_buff member frame_sz. (Björn)
Add API to query the DMA address of a frame. (Maxim)
Do DMA sync for CPU till the end of the frame to handle
possible growth (frame_sz). (Maxim)
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200520192103.355233-6-bjorn.topel@gmail.com
Move the XSK_NEXT_PG_CONTIG_{MASK,SHIFT}, and
XDP_UMEM_USES_NEED_WAKEUP defines from xdp_sock.h to the AF_XDP
internal xsk.h file. Also, start using the BIT{,_ULL} macro instead of
explicit shifts.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200520192103.355233-5-bjorn.topel@gmail.com
Move the AF_XDP zero-copy driver interface to its own include file
called xdp_sock_drv.h. This, hopefully, will make it more clear for
NIC driver implementors to know what functions to use for zero-copy
support.
v4->v5: Fix -Wmissing-prototypes by include header file. (Jakub)
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200520192103.355233-4-bjorn.topel@gmail.com
The XSKMAP is partly implemented by net/xdp/xsk.c. Move xskmap.c from
kernel/bpf/ to net/xdp/, which is the logical place for AF_XDP related
code. Also, move AF_XDP struct definitions, and function declarations
only used by AF_XDP internals into net/xdp/xsk.h.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200520192103.355233-3-bjorn.topel@gmail.com
Remove the unnecessary member of address in struct xdp_umem as it is
only used during the umem registration. No need to carry this around
as it is not used during run-time nor when unregistering the umem.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/bpf/1588599232-24897-3-git-send-email-magnus.karlsson@intel.com
Change two variables names so that it is clearer what they
represent. The first one is xsk_list that in fact only contains the
list of AF_XDP sockets with a Tx component. Change this to xsk_tx_list
for improved clarity. The second variable is size in the ring
structure. One might think that this is the size of the ring, but it
is in fact the size of the umem, copied into the ring structure to
improve performance. Rename this variable umem_size to avoid any
confusion.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/bpf/1588599232-24897-2-git-send-email-magnus.karlsson@intel.com
Add a check that the headroom cannot be larger than the available
space in the chunk. In the current code, a malicious user can set the
headroom to a value larger than the chunk size minus the fixed XDP
headroom. That way packets with a length larger than the supported
size in the umem could get accepted and result in an out-of-bounds
write.
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Reported-by: Bui Quang Minh <minhquangbui99@gmail.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=207225
Link: https://lore.kernel.org/bpf/1586849715-23490-1-git-send-email-magnus.karlsson@intel.com
first_len is the remainder of the first page we're copying.
If this size is larger, then out of page boundary write will
otherwise happen.
Fixes: c05cd36458 ("xsk: add support to allow unaligned chunk placement")
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1585813930-19712-1-git-send-email-lirongqing@baidu.com
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The commit 4b638f13ba ("xsk: Eliminate the RX batch size")
introduced a much more lazy way of updating the global consumer
pointers from the kernel side, by only doing so when running out of
entries in the fill or Tx rings (the rings consumed by the
kernel). This can result in a deadlock with the user application if
the kernel requires more than one entry to proceed and the application
cannot put these entries in the fill ring because the kernel has not
updated the global consumer pointer since the ring is not empty.
Fix this by publishing the local kernel side consumer pointer whenever
we have completed Rx or Tx processing in the kernel. This way, user
space will have an up-to-date view of the consumer pointers whenever it
gets to execute in the one core case (application and driver on the
same core), or after a certain number of packets have been processed
in the two core case (application and driver on different cores).
A side effect of this patch is that the one core case gets better
performance, but the two core case gets worse. The reason that the one
core case improves is that updating the global consumer pointer is
relatively cheap since the application by definition is not running
when the kernel is (they are on the same core) and it is beneficial
for the application, once it gets to run, to have pointers that are
as up to date as possible since it then can operate on more packets
and buffers. In the two core case, the most important performance
aspect is to minimize the number of accesses to the global pointers
since they are shared between two cores and bounces between the caches
of those cores. This patch results in more updates to global state,
which means lower performance in the two core case.
Fixes: 4b638f13ba ("xsk: Eliminate the RX batch size")
Reported-by: Ryan Goodfellow <rgoodfel@isi.edu>
Reported-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Link: https://lore.kernel.org/bpf/1581348432-6747-1-git-send-email-magnus.karlsson@intel.com
In order to provide a clearer, more symmetric API for pinning and
unpinning DMA pages. This way, pin_user_pages*() calls match up with
unpin_user_pages*() calls, and the API is a lot closer to being
self-explanatory.
Link: http://lkml.kernel.org/r/20200107224558.2362728-23-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Convert net/xdp to use the new pin_longterm_pages() call, which sets
FOLL_PIN. Setting FOLL_PIN is now required for code that requires
tracking of pinned pages.
In partial anticipation of this work, the net/xdp code was already calling
put_user_page() instead of put_page(). Therefore, in order to convert
from the get_user_pages()/put_page() model, to the
pin_user_pages()/put_user_page() model, the only change required here is
to change get_user_pages() to pin_user_pages().
Link: http://lkml.kernel.org/r/20200107224558.2362728-18-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
XDP sockets use the default implementation of struct sock's
sk_data_ready callback, which is sock_def_readable(). This function
is called in the XDP socket fast-path, and involves a retpoline. By
letting sock_def_readable() have external linkage, and being called
directly, the retpoline can be avoided.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200120092917.13949-1-bjorn.topel@gmail.com
When registering a umem area that is sufficiently large (>1G on an
x86), kmalloc cannot be used to allocate one of the internal data
structures, as the size requested gets too large. Use kvmalloc instead
that falls back on vmalloc if the allocation is too large for kmalloc.
Also add accounting for this structure as it is triggered by a user
space action (the XDP_UMEM_REG setsockopt) and it is by far the
largest structure of kernel allocated memory in xsk.
Reported-by: Ryan Goodfellow <rgoodfel@isi.edu>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/bpf/1578995365-7050-1-git-send-email-magnus.karlsson@intel.com
Daniel Borkmann says:
====================
pull-request: bpf-next 2019-12-27
The following pull-request contains BPF updates for your *net-next* tree.
We've added 127 non-merge commits during the last 17 day(s) which contain
a total of 110 files changed, 6901 insertions(+), 2721 deletions(-).
There are three merge conflicts. Conflicts and resolution looks as follows:
1) Merge conflict in net/bpf/test_run.c:
There was a tree-wide cleanup c593642c8b ("treewide: Use sizeof_field() macro")
which gets in the way with b590cb5f80 ("bpf: Switch to offsetofend in
BPF_PROG_TEST_RUN"):
<<<<<<< HEAD
if (!range_is_zero(__skb, offsetof(struct __sk_buff, priority) +
sizeof_field(struct __sk_buff, priority),
=======
if (!range_is_zero(__skb, offsetofend(struct __sk_buff, priority),
>>>>>>> 7c8dce4b16
There are a few occasions that look similar to this. Always take the chunk with
offsetofend(). Note that there is one where the fields differ in here:
<<<<<<< HEAD
if (!range_is_zero(__skb, offsetof(struct __sk_buff, tstamp) +
sizeof_field(struct __sk_buff, tstamp),
=======
if (!range_is_zero(__skb, offsetofend(struct __sk_buff, gso_segs),
>>>>>>> 7c8dce4b16
Just take the one with offsetofend() /and/ gso_segs. Latter is correct due to
850a88cc40 ("bpf: Expose __sk_buff wire_len/gso_segs to BPF_PROG_TEST_RUN").
2) Merge conflict in arch/riscv/net/bpf_jit_comp.c:
(I'm keeping Bjorn in Cc here for a double-check in case I got it wrong.)
<<<<<<< HEAD
if (is_13b_check(off, insn))
return -1;
emit(rv_blt(tcc, RV_REG_ZERO, off >> 1), ctx);
=======
emit_branch(BPF_JSLT, RV_REG_T1, RV_REG_ZERO, off, ctx);
>>>>>>> 7c8dce4b16
Result should look like:
emit_branch(BPF_JSLT, tcc, RV_REG_ZERO, off, ctx);
3) Merge conflict in arch/riscv/include/asm/pgtable.h:
<<<<<<< HEAD
=======
#define VMALLOC_SIZE (KERN_VIRT_SIZE >> 1)
#define VMALLOC_END (PAGE_OFFSET - 1)
#define VMALLOC_START (PAGE_OFFSET - VMALLOC_SIZE)
#define BPF_JIT_REGION_SIZE (SZ_128M)
#define BPF_JIT_REGION_START (PAGE_OFFSET - BPF_JIT_REGION_SIZE)
#define BPF_JIT_REGION_END (VMALLOC_END)
/*
* Roughly size the vmemmap space to be large enough to fit enough
* struct pages to map half the virtual address space. Then
* position vmemmap directly below the VMALLOC region.
*/
#define VMEMMAP_SHIFT \
(CONFIG_VA_BITS - PAGE_SHIFT - 1 + STRUCT_PAGE_MAX_SHIFT)
#define VMEMMAP_SIZE BIT(VMEMMAP_SHIFT)
#define VMEMMAP_END (VMALLOC_START - 1)
#define VMEMMAP_START (VMALLOC_START - VMEMMAP_SIZE)
#define vmemmap ((struct page *)VMEMMAP_START)
>>>>>>> 7c8dce4b16
Only take the BPF_* defines from there and move them higher up in the
same file. Remove the rest from the chunk. The VMALLOC_* etc defines
got moved via 01f52e16b8 ("riscv: define vmemmap before pfn_to_page
calls"). Result:
[...]
#define __S101 PAGE_READ_EXEC
#define __S110 PAGE_SHARED_EXEC
#define __S111 PAGE_SHARED_EXEC
#define VMALLOC_SIZE (KERN_VIRT_SIZE >> 1)
#define VMALLOC_END (PAGE_OFFSET - 1)
#define VMALLOC_START (PAGE_OFFSET - VMALLOC_SIZE)
#define BPF_JIT_REGION_SIZE (SZ_128M)
#define BPF_JIT_REGION_START (PAGE_OFFSET - BPF_JIT_REGION_SIZE)
#define BPF_JIT_REGION_END (VMALLOC_END)
/*
* Roughly size the vmemmap space to be large enough to fit enough
* struct pages to map half the virtual address space. Then
* position vmemmap directly below the VMALLOC region.
*/
#define VMEMMAP_SHIFT \
(CONFIG_VA_BITS - PAGE_SHIFT - 1 + STRUCT_PAGE_MAX_SHIFT)
#define VMEMMAP_SIZE BIT(VMEMMAP_SHIFT)
#define VMEMMAP_END (VMALLOC_START - 1)
#define VMEMMAP_START (VMALLOC_START - VMEMMAP_SIZE)
[...]
Let me know if there are any other issues.
Anyway, the main changes are:
1) Extend bpftool to produce a struct (aka "skeleton") tailored and specific
to a provided BPF object file. This provides an alternative, simplified API
compared to standard libbpf interaction. Also, add libbpf extern variable
resolution for .kconfig section to import Kconfig data, from Andrii Nakryiko.
2) Add BPF dispatcher for XDP which is a mechanism to avoid indirect calls by
generating a branch funnel as discussed back in bpfconf'19 at LSF/MM. Also,
add various BPF riscv JIT improvements, from Björn Töpel.
3) Extend bpftool to allow matching BPF programs and maps by name,
from Paul Chaignon.
4) Support for replacing cgroup BPF programs attached with BPF_F_ALLOW_MULTI
flag for allowing updates without service interruption, from Andrey Ignatov.
5) Cleanup and simplification of ring access functions for AF_XDP with a
bonus of 0-5% performance improvement, from Magnus Karlsson.
6) Enable BPF JITs for x86-64 and arm64 by default. Also, final version of
audit support for BPF, from Daniel Borkmann and latter with Jiri Olsa.
7) Move and extend test_select_reuseport into BPF program tests under
BPF selftests, from Jakub Sitnicki.
8) Various BPF sample improvements for xdpsock for customizing parameters
to set up and benchmark AF_XDP, from Jay Jayatheerthan.
9) Improve libbpf to provide a ulimit hint on permission denied errors.
Also change XDP sample programs to attach in driver mode by default,
from Toke Høiland-Jørgensen.
10) Extend BPF test infrastructure to allow changing skb mark from tc BPF
programs, from Nikita V. Shirokov.
11) Optimize prologue code sequence in BPF arm32 JIT, from Russell King.
12) Fix xdp_redirect_cpu BPF sample to manually attach to tracepoints after
libbpf conversion, from Jesper Dangaard Brouer.
13) Minor misc improvements from various others.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add comments on how the ring access functions are named and how they
are supposed to be used for producers and consumers. The functions are
also reordered so that the consumer functions are in the beginning and
the producer functions in the end, for easier reference. Put this in a
separate patch as the diff might look a little odd, but no
functionality has changed in this patch.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-12-git-send-email-magnus.karlsson@intel.com
There are two unnecessary READ_ONCE of descriptor data. These are not
needed since the data is written by the producer before it signals
that the data is available by incrementing the producer pointer. As the
access to this producer pointer is serialized and the consumer always
reads the descriptor after it has read and synchronized with the
producer counter, the write of the descriptor will have fully
completed and it does not matter if the consumer has any read tearing.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-11-git-send-email-magnus.karlsson@intel.com
Change the name of xsk_umem_discard_addr to xsk_umem_release_addr to
better reflect the new naming of the AF_XDP queue manipulation
functions. As this functions is used by drivers implementing support
for AF_XDP zero-copy, it requires a name change to these drivers. The
function xsk_umem_release_addr_rq has also changed name in the same
fashion.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-10-git-send-email-magnus.karlsson@intel.com
Change the names of the validation functions to better reflect what
they are doing. The uppermost ones are reading entries from the rings
and only the bottom ones validate entries. So xskq_cons_read_ is a
better prefix name.
Also change the xskq_cons_read_ functions to return a bool
as the the descriptor or address is already returned by reference
in the parameters. Everyone is using the return value as a bool
anyway.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-9-git-send-email-magnus.karlsson@intel.com
Simplify and refactor consumer ring functions. The consumer first
"peeks" to find descriptors or addresses that are available to
read from the ring, then reads them and finally "releases" these
descriptors once it is done. The two local variables cons_tail
and cons_head are turned into one single variable called
cached_cons. cached_tail referred to the cached value of the
global consumer pointer and will be stored in cached_cons. For
cached_head, we just use cached_prod instead as it was not used
for a consumer queue before. It also better reflects what it
really is now: a cached copy of the producer pointer.
The names of the functions are also renamed in the same manner as
the producer functions. The new functions are called xskq_cons_
followed by what it does.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-8-git-send-email-magnus.karlsson@intel.com
At this point, there are no users of the functions xskq_nb_avail and
xskq_nb_free that take any other number of entries argument than 1, so
let us get rid of the second argument that takes the number of
entries.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-7-git-send-email-magnus.karlsson@intel.com
In the xsk consumer ring code there is a variable called RX_BATCH_SIZE
that dictates the minimum number of entries that we try to grab from
the fill and Tx rings. In fact, the code always try to grab the
maximum amount of entries from these rings. The only thing this
variable does is to throw an error if there is less than 16 (as it is
defined) entries on the ring. There is no reason to do this and it
will just lead to weird behavior from user space's point of view. So
eliminate this variable.
With this change, we will be able to simplify the xskq_nb_free and
xskq_nb_avail code in the next commit.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-6-git-send-email-magnus.karlsson@intel.com
Adopt the naming of the producer ring access functions to have a
similar naming convention as the functions in libbpf, but adapted to
the kernel. You first reserve a number of entries that you later
submit to the global state of the ring. This is much clearer, IMO,
than the one that was in the kernel part. Once renamed, we also
discover that two functions are actually the same, so remove one of
them. Some of the primitive ring submission operations are also the
same so break these out into __xskq_prod_submit that the upper level
ring access functions can use.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-5-git-send-email-magnus.karlsson@intel.com
Currently, the xsk ring code has two cached producer pointers:
prod_head and prod_tail. This patch consolidates these two into a
single one called cached_prod to make the code simpler and easier to
maintain. This will be in line with the user space part of the the
code found in libbpf, that only uses a single cached pointer.
The Rx path only uses the two top level functions
xskq_produce_batch_desc and xskq_produce_flush_desc and they both use
prod_head and never prod_tail. So just move them over to
cached_prod.
The Tx XDP_DRV path uses xskq_produce_addr_lazy and
xskq_produce_flush_addr_n and unnecessarily operates on both prod_tail
and prod_head, so move them over to just use cached_prod by skipping
the intermediate step of updating prod_tail.
The Tx path in XDP_SKB mode uses xskq_reserve_addr and
xskq_produce_addr. They currently use both cached pointers, but we can
operate on the global producer pointer in xskq_produce_addr since it
has to be updated anyway, thus eliminating the use of both cached
pointers. We can also remove the xskq_nb_free in xskq_produce_addr
since it is already called in xskq_reserve_addr. No need to do it
twice.
When there is only one cached producer pointer, we can also simplify
xskq_nb_free by removing one argument.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-4-git-send-email-magnus.karlsson@intel.com
In order to set the correct return flags for poll, the xsk code has to
check if the Rx queue is empty and if the Tx queue is full. This code
was unnecessarily large and complex as it used the functions that are
used to update the local state from the global state (xskq_nb_free and
xskq_nb_avail). Since we are not doing this nor updating any data
dependent on this state, we can simplify the functions. Another
benefit from this is that we can also simplify the xskq_nb_free and
xskq_nb_avail functions in a later commit.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-3-git-send-email-magnus.karlsson@intel.com
The lazy update threshold was introduced to keep the producer and
consumer some distance apart in the completion ring. This was
important in the beginning of the development of AF_XDP as the ring
format as that point in time was very sensitive to the producer and
consumer being on the same cache line. This is not the case
anymore as the current ring format does not degrade in any noticeable
way when this happens. Moreover, this threshold makes it impossible
to run the system with rings that have less than 128 entries.
So let us remove this threshold and just get one entry from the ring
as in all other functions. This will enable us to remove this function
in a later commit. Note that xskq_produce_addr_lazy followed by
xskq_produce_flush_addr_n are still not the same function as
xskq_produce_addr() as it operates on another cached pointer.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/1576759171-28550-2-git-send-email-magnus.karlsson@intel.com
The xskmap flush list is used to track entries that need to flushed
from via the xdp_do_flush_map() function. This list used to be
per-map, but there is really no reason for that. Instead make the
flush list global for all xskmaps, which simplifies __xsk_map_flush()
and xsk_map_alloc().
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/bpf/20191219061006.21980-5-bjorn.topel@gmail.com
The XSK wakeup callback in drivers makes some sanity checks before
triggering NAPI. However, some configuration changes may occur during
this function that affect the result of those checks. For example, the
interface can go down, and all the resources will be destroyed after the
checks in the wakeup function, but before it attempts to use these
resources. Wrap this callback in rcu_read_lock to allow driver to
synchronize_rcu before actually destroying the resources.
xsk_wakeup is a new function that encapsulates calling ndo_xsk_wakeup
wrapped into the RCU lock. After this commit, xsk_poll starts using
xsk_wakeup and checks xs->zc instead of ndo_xsk_wakeup != NULL to decide
ndo_xsk_wakeup should be called. It also fixes a bug introduced with the
need_wakeup feature: a non-zero-copy socket may be used with a driver
supporting zero-copy, and in this case ndo_xsk_wakeup should not be
called, so the xs->zc check is the correct one.
Fixes: 77cd0d7b3f ("xsk: add support for need_wakeup flag in AF_XDP rings")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20191217162023.16011-2-maximmi@mellanox.com
xsk_poll() is defined as returning 'unsigned int' but the
.poll method is declared as returning '__poll_t', a bitwise type.
Fix this by using the proper return type and using the EPOLL
constants instead of the POLL ones, as required for __poll_t.
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/20191120001042.30830-1-luc.vanoostenryck@gmail.com
Alexei Starovoitov says:
====================
pull-request: bpf-next 2019-11-02
The following pull-request contains BPF updates for your *net-next* tree.
We've added 30 non-merge commits during the last 7 day(s) which contain
a total of 41 files changed, 1864 insertions(+), 474 deletions(-).
The main changes are:
1) Fix long standing user vs kernel access issue by introducing
bpf_probe_read_user() and bpf_probe_read_kernel() helpers, from Daniel.
2) Accelerated xskmap lookup, from Björn and Maciej.
3) Support for automatic map pinning in libbpf, from Toke.
4) Cleanup of BTF-enabled raw tracepoints, from Alexei.
5) Various fixes to libbpf and selftests.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
In this commit the XSKMAP entry lookup function used by the XDP
redirect code is moved from the xskmap.c file to the xdp_sock.h
header, so the lookup can be inlined from, e.g., the
bpf_xdp_redirect_map() function.
Further the __xsk_map_redirect() and __xsk_map_flush() is moved to the
xsk.c, which lets the compiler inline the xsk_rcv() and xsk_flush()
functions.
Finally, all the XDP socket functions were moved from linux/bpf.h to
net/xdp_sock.h, where most of the XDP sockets functions are anyway.
This yields a ~2% performance boost for the xdpsock "rx_drop"
scenario.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20191101110346.15004-4-bjorn.topel@gmail.com
Having Rx-only AF_XDP sockets can potentially lead to a crash in the
system by a NULL pointer dereference in xsk_umem_consume_tx(). This
function iterates through a list of all sockets tied to a umem and
checks if there are any packets to send on the Tx ring. Rx-only
sockets do not have a Tx ring, so this will cause a NULL pointer
dereference. This will happen if you have registered one or more
Rx-only sockets to a umem and the driver is checking the Tx ring even
on Rx, or if the XDP_SHARED_UMEM mode is used and there is a mix of
Rx-only and other sockets tied to the same umem.
Fixed by only putting sockets with a Tx component on the list that
xsk_umem_consume_tx() iterates over.
Fixes: ac98d8aab6 ("xsk: wire upp Tx zero-copy functions")
Reported-by: Kal Cutter Conley <kal.conley@dectris.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/bpf/1571645818-16244-1-git-send-email-magnus.karlsson@intel.com
Fixes a crash in poll() when an AF_XDP socket is opened in copy mode
and the bound device does not have ndo_xsk_wakeup defined. Avoid
trying to call the non-existing ndo and instead call the internal xsk
sendmsg function to send packets in the same way (from the
application's point of view) as calling sendmsg() in any mode or
poll() in zero-copy mode would have done. The application should
behave in the same way independent on if zero-copy mode or copy mode
is used.
Fixes: 77cd0d7b3f ("xsk: add support for need_wakeup flag in AF_XDP rings")
Reported-by: syzbot+a5765ed8cdb1cca4d249@syzkaller.appspotmail.com
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1569997919-11541-1-git-send-email-magnus.karlsson@intel.com
Pull networking fixes from David Miller:
1) Sanity check URB networking device parameters to avoid divide by
zero, from Oliver Neukum.
2) Disable global multicast filter in NCSI, otherwise LLDP and IPV6
don't work properly. Longer term this needs a better fix tho. From
Vijay Khemka.
3) Small fixes to selftests (use ping when ping6 is not present, etc.)
from David Ahern.
4) Bring back rt_uses_gateway member of struct rtable, it's semantics
were not well understood and trying to remove it broke things. From
David Ahern.
5) Move usbnet snaity checking, ignore endpoints with invalid
wMaxPacketSize. From Bjørn Mork.
6) Missing Kconfig deps for sja1105 driver, from Mao Wenan.
7) Various small fixes to the mlx5 DR steering code, from Alaa Hleihel,
Alex Vesker, and Yevgeny Kliteynik
8) Missing CAP_NET_RAW checks in various places, from Ori Nimron.
9) Fix crash when removing sch_cbs entry while offloading is enabled,
from Vinicius Costa Gomes.
10) Signedness bug fixes, generally in looking at the result given by
of_get_phy_mode() and friends. From Dan Crapenter.
11) Disable preemption around BPF_PROG_RUN() calls, from Eric Dumazet.
12) Don't create VRF ipv6 rules if ipv6 is disabled, from David Ahern.
13) Fix quantization code in tcp_bbr, from Kevin Yang.
* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (127 commits)
net: tap: clean up an indentation issue
nfp: abm: fix memory leak in nfp_abm_u32_knode_replace
tcp: better handle TCP_USER_TIMEOUT in SYN_SENT state
sk_buff: drop all skb extensions on free and skb scrubbing
tcp_bbr: fix quantization code to not raise cwnd if not probing bandwidth
mlxsw: spectrum_flower: Fail in case user specifies multiple mirror actions
Documentation: Clarify trap's description
mlxsw: spectrum: Clear VLAN filters during port initialization
net: ena: clean up indentation issue
NFC: st95hf: clean up indentation issue
net: phy: micrel: add Asym Pause workaround for KSZ9021
net: socionext: ave: Avoid using netdev_err() before calling register_netdev()
ptp: correctly disable flags on old ioctls
lib: dimlib: fix help text typos
net: dsa: microchip: Always set regmap stride to 1
nfp: flower: fix memory leak in nfp_flower_spawn_vnic_reprs
nfp: flower: prevent memory leak in nfp_flower_spawn_phy_reprs
net/sched: Set default of CONFIG_NET_TC_SKB_EXT to N
vrf: Do not attempt to create IPv6 mcast rule if IPv6 is disabled
net: sched: sch_sfb: don't call qdisc_put() while holding tree lock
...
For pages that were retained via get_user_pages*(), release those pages
via the new put_user_page*() routines, instead of via put_page() or
release_pages().
This is part a tree-wide conversion, as described in fc1d8e7cca ("mm:
introduce put_user_page*(), placeholder versions").
Link: http://lkml.kernel.org/r/20190724044537.10458-4-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Magnus Karlsson <magnus.karlsson@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "Make working with compound pages easier", v2.
These three patches add three helpers and convert the appropriate
places to use them.
This patch (of 3):
It's unnecessarily hard to find out the size of a potentially huge page.
Replace 'PAGE_SIZE << compound_order(page)' with page_size(page).
Link: http://lkml.kernel.org/r/20190721104612.19120-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch removes the 64B alignment of the UMEM headroom. There is
really no reason for it, and having a headroom less than 64B should be
valid.
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Daniel Borkmann says:
====================
The following pull-request contains BPF updates for your *net-next* tree.
The main changes are:
1) Add the ability to use unaligned chunks in the AF_XDP umem. By
relaxing where the chunks can be placed, it allows to use an
arbitrary buffer size and place whenever there is a free
address in the umem. Helps more seamless DPDK AF_XDP driver
integration. Support for i40e, ixgbe and mlx5e, from Kevin and
Maxim.
2) Addition of a wakeup flag for AF_XDP tx and fill rings so the
application can wake up the kernel for rx/tx processing which
avoids busy-spinning of the latter, useful when app and driver
is located on the same core. Support for i40e, ixgbe and mlx5e,
from Magnus and Maxim.
3) bpftool fixes for printf()-like functions so compiler can actually
enforce checks, bpftool build system improvements for custom output
directories, and addition of 'bpftool map freeze' command, from Quentin.
4) Support attaching/detaching XDP programs from 'bpftool net' command,
from Daniel.
5) Automatic xskmap cleanup when AF_XDP socket is released, and several
barrier/{read,write}_once fixes in AF_XDP code, from Björn.
6) Relicense of bpf_helpers.h/bpf_endian.h for future libbpf
inclusion as well as libbpf versioning improvements, from Andrii.
7) Several new BPF kselftests for verifier precision tracking, from Alexei.
8) Several BPF kselftest fixes wrt endianess to run on s390x, from Ilya.
9) And more BPF kselftest improvements all over the place, from Stanislav.
10) Add simple BPF map op cache for nfp driver to batch dumps, from Jakub.
11) AF_XDP socket umem mapping improvements for 32bit archs, from Ivan.
12) Add BPF-to-BPF call and BTF line info support for s390x JIT, from Yauheni.
13) Small optimization in arm64 JIT to spare 1 insns for BPF_MOD, from Jerin.
14) Fix an error check in bpf_tcp_gen_syncookie() helper, from Petar.
15) Various minor fixes and cleanups, from Nathan, Masahiro, Masanari,
Peter, Wei, Yue.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When accessing the members of an XDP socket, the control mutex should
be held. This commit fixes that.
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Fixes: a36b38aa2a ("xsk: add sock_diag interface for AF_XDP")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Prior the state variable was introduced by Ilya, the dev member was
used to determine whether the socket was bound or not. However, when
dev was read, proper SMP barriers and READ_ONCE were missing. In order
to address the missing barriers and READ_ONCE, we start using the
state variable as a point of synchronization. The state member
read/write is paired with proper SMP barriers, and from this follows
that the members described above does not need READ_ONCE if used in
conjunction with state check.
In all syscalls and the xsk_rcv path we check if state is
XSK_BOUND. If that is the case we do a SMP read barrier, and this
implies that the dev, umem and all rings are correctly setup. Note
that no READ_ONCE are needed for these variable if used when state is
XSK_BOUND (plus the read barrier).
To summarize: The members struct xdp_sock members dev, queue_id, umem,
fq, cq, tx, rx, and state were read lock-less, with incorrect barriers
and missing {READ, WRITE}_ONCE. Now, umem, fq, cq, tx, rx, and state
are read lock-less. When these members are updated, WRITE_ONCE is
used. When read, READ_ONCE are only used when read outside the control
mutex (e.g. mmap) or, not synchronized with the state member
(XSK_BOUND plus smp_rmb())
Note that dev and queue_id do not need a WRITE_ONCE or READ_ONCE, due
to the introduce state synchronization (XSK_BOUND plus smp_rmb()).
Introducing the state check also fixes a race, found by syzcaller, in
xsk_poll() where umem could be accessed when stale.
Suggested-by: Hillf Danton <hdanton@sina.com>
Reported-by: syzbot+c82697e3043781e08802@syzkaller.appspotmail.com
Fixes: 77cd0d7b3f ("xsk: add support for need_wakeup flag in AF_XDP rings")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The umem member of struct xdp_sock is read outside of the control
mutex, in the mmap implementation, and needs a WRITE_ONCE to avoid
potential store-tearing.
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Fixes: 423f38329d ("xsk: add umem fill queue support and mmap")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Use WRITE_ONCE when doing the store of tx, rx, fq, and cq, to avoid
potential store-tearing. These members are read outside of the control
mutex in the mmap implementation.
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Fixes: 37b076933a ("xsk: add missing write- and data-dependency barrier")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently, addresses are chunk size aligned. This means, we are very
restricted in terms of where we can place chunk within the umem. For
example, if we have a chunk size of 2k, then our chunks can only be placed
at 0,2k,4k,6k,8k... and so on (ie. every 2k starting from 0).
This patch introduces the ability to use unaligned chunks. With these
changes, we are no longer bound to having to place chunks at a 2k (or
whatever your chunk size is) interval. Since we are no longer dealing with
aligned chunks, they can now cross page boundaries. Checks for page
contiguity have been added in order to keep track of which pages are
followed by a physically contiguous page.
Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For 64-bit there is no reason to use vmap/vunmap, so use page_address
as it was initially. For 32 bits, in some apps, like in samples
xdpsock_user.c when number of pgs in use is quite big, the kmap
memory can be not enough, despite on this, kmap looks like is
deprecated in such cases as it can block and should be used rather
for dynamic mm.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Fix mem leak caused by missed unpin routine for umem pages.
Fixes: 8aef7340ae ("xsk: introduce xdp_umem_page")
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When an AF_XDP socket is released/closed the XSKMAP still holds a
reference to the socket in a "released" state. The socket will still
use the netdev queue resource, and block newly created sockets from
attaching to that queue, but no user application can access the
fill/complete/rx/tx queues. This results in that all applications need
to explicitly clear the map entry from the old "zombie state"
socket. This should be done automatically.
In this patch, the sockets tracks, and have a reference to, which maps
it resides in. When the socket is released, it will remove itself from
all maps.
Suggested-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds support for a new flag called need_wakeup in the
AF_XDP Tx and fill rings. When this flag is set, it means that the
application has to explicitly wake up the kernel Rx (for the bit in
the fill ring) or kernel Tx (for bit in the Tx ring) processing by
issuing a syscall. Poll() can wake up both depending on the flags
submitted and sendto() will wake up tx processing only.
The main reason for introducing this new flag is to be able to
efficiently support the case when application and driver is executing
on the same core. Previously, the driver was just busy-spinning on the
fill ring if it ran out of buffers in the HW and there were none on
the fill ring. This approach works when the application is running on
another core as it can replenish the fill ring while the driver is
busy-spinning. Though, this is a lousy approach if both of them are
running on the same core as the probability of the fill ring getting
more entries when the driver is busy-spinning is zero. With this new
feature the driver now sets the need_wakeup flag and returns to the
application. The application can then replenish the fill queue and
then explicitly wake up the Rx processing in the kernel using the
syscall poll(). For Tx, the flag is only set to one if the driver has
no outstanding Tx completion interrupts. If it has some, the flag is
zero as it will be woken up by a completion interrupt anyway.
As a nice side effect, this new flag also improves the performance of
the case where application and driver are running on two different
cores as it reduces the number of syscalls to the kernel. The kernel
tells user space if it needs to be woken up by a syscall, and this
eliminates many of the syscalls.
This flag needs some simple driver support. If the driver does not
support this, the Rx flag is always zero and the Tx flag is always
one. This makes any application relying on this feature default to the
old behaviour of not requiring any syscalls in the Rx path and always
having to call sendto() in the Tx path.
For backwards compatibility reasons, this feature has to be explicitly
turned on using a new bind flag (XDP_USE_NEED_WAKEUP). I recommend
that you always turn it on as it so far always have had a positive
performance impact.
The name and inspiration of the flag has been taken from io_uring by
Jens Axboe. Details about this feature in io_uring can be found in
http://kernel.dk/io_uring.pdf, section 8.3.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit replaces ndo_xsk_async_xmit with ndo_xsk_wakeup. This new
ndo provides the same functionality as before but with the addition of
a new flags field that is used to specifiy if Rx, Tx or both should be
woken up. The previous ndo only woke up Tx, as implied by the
name. The i40e and ixgbe drivers (which are all the supported ones)
are updated with this new interface.
This new ndo will be used by the new need_wakeup functionality of XDP
sockets that need to be able to wake up both Rx and Tx driver
processing.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Use kmap instead of page_address as it's not always in low memory.
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
There are 2 call chains:
a) xsk_bind --> xdp_umem_assign_dev
b) unregister_netdevice_queue --> xsk_notifier
with the following locking order:
a) xs->mutex --> rtnl_lock
b) rtnl_lock --> xdp.lock --> xs->mutex
Different order of taking 'xs->mutex' and 'rtnl_lock' could produce a
deadlock here. Fix that by moving the 'rtnl_lock' before 'xs->lock' in
the bind call chain (a).
Reported-by: syzbot+bf64ec93de836d7f4c2c@syzkaller.appspotmail.com
Fixes: 455302d1c9 ("xdp: fix hang while unregistering device bound to xdp socket")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Completion queue address reservation could not be undone.
In case of bad 'queue_id' or skb allocation failure, reserved entry
will be leaked reducing the total capacity of completion queue.
Fix that by moving reservation to the point where failure is not
possible. Additionally, 'queue_id' checking moved out from the loop
since there is no point to check it there.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: William Tu <u9012063@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Unlike driver mode, generic xdp receive could be triggered
by different threads on different CPU cores at the same time
leading to the fill and rx queue breakage. For example, this
could happen while sending packets from two processes to the
first interface of veth pair while the second part of it is
open with AF_XDP socket.
Need to take a lock for each generic receive to avoid race.
Fixes: c497176cb2 ("xsk: add Rx receive functions and poll support")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Tested-by: William Tu <u9012063@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Device that bound to XDP socket will not have zero refcount until the
userspace application will not close it. This leads to hang inside
'netdev_wait_allrefs()' if device unregistering requested:
# ip link del p1
< hang on recvmsg on netlink socket >
# ps -x | grep ip
5126 pts/0 D+ 0:00 ip link del p1
# journalctl -b
Jun 05 07:19:16 kernel:
unregister_netdevice: waiting for p1 to become free. Usage count = 1
Jun 05 07:19:27 kernel:
unregister_netdevice: waiting for p1 to become free. Usage count = 1
...
Fix that by implementing NETDEV_UNREGISTER event notification handler
to properly clean up all the resources and unref device.
This should also allow socket killing via ss(8) utility.
Fixes: 965a990984 ("xsk: add support for bind for Rx")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Device pointer stored in umem regardless of zero-copy mode,
so we heed to hold the device in all cases.
Fixes: c9b47cc1fa ("xsk: fix bug when trying to use both copy and zero-copy on one queue id")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Some drivers want to access the data transmitted in order to implement
acceleration features of the NICs. It is also useful in AF_XDP TX flow.
Change the xsk_umem_consume_tx API to return the whole xdp_desc, that
contains the data pointer, length and DMA address, instead of only the
latter two. Adapt the implementation of i40e and ixgbe to this change.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Make it possible for the application to determine whether the AF_XDP
socket is running in zero-copy mode. To achieve this, add a new
getsockopt option XDP_OPTIONS that returns flags. The only flag
supported for now is the zero-copy mode indicator.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add a function that checks whether the Fill Ring has the specified
amount of descriptors available. It will be useful for mlx5e that wants
to check in advance, whether it can allocate a bulk of RX descriptors,
to get the best performance.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Clang warns:
In file included from net/xdp/xsk_queue.c:10:
net/xdp/xsk_queue.h:292:2: warning: expression result unused
[-Wunused-value]
WRITE_ONCE(q->ring->producer, q->prod_tail);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/compiler.h:284:6: note: expanded from macro 'WRITE_ONCE'
__u.__val; \
~~~ ^~~~~
1 warning generated.
The q->prod_tail assignment has a comma at the end, not a semi-colon.
Fix that so clang no longer warns and everything works as expected.
Fixes: c497176cb2 ("xsk: add Rx receive functions and poll support")
Link: https://github.com/ClangBuiltLinux/linux/issues/544
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
We should not call 'ndo_bpf()' or 'dev_put()' with NULL argument.
Fixes: c9b47cc1fa ("xsk: fix bug when trying to use both copy and zero-copy on one queue id")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add SPDX license identifiers to all Make/Kconfig files which:
- Have no license information of any form
These files fall under the project license, GPL v2 only. The resulting SPDX
license identifier is:
GPL-2.0-only
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pach series "Add FOLL_LONGTERM to GUP fast and use it".
HFI1, qib, and mthca, use get_user_pages_fast() due to its performance
advantages. These pages can be held for a significant time. But
get_user_pages_fast() does not protect against mapping FS DAX pages.
Introduce FOLL_LONGTERM and use this flag in get_user_pages_fast() which
retains the performance while also adding the FS DAX checks. XDP has also
shown interest in using this functionality.[1]
In addition we change get_user_pages() to use the new FOLL_LONGTERM flag
and remove the specialized get_user_pages_longterm call.
[1] https://lkml.org/lkml/2019/3/19/939
"longterm" is a relative thing and at this point is probably a misnomer.
This is really flagging a pin which is going to be given to hardware and
can't move. I've thought of a couple of alternative names but I think we
have to settle on if we are going to use FL_LAYOUT or something else to
solve the "longterm" problem. Then I think we can change the flag to a
better name.
Secondly, it depends on how often you are registering memory. I have
spoken with some RDMA users who consider MR in the performance path...
For the overall application performance. I don't have the numbers as the
tests for HFI1 were done a long time ago. But there was a significant
advantage. Some of which is probably due to the fact that you don't have
to hold mmap_sem.
Finally, architecturally I think it would be good for everyone to use
*_fast. There are patches submitted to the RDMA list which would allow
the use of *_fast (they reworking the use of mmap_sem) and as soon as they
are accepted I'll submit a patch to convert the RDMA core as well. Also
to this point others are looking to use *_fast.
As an aside, Jasons pointed out in my previous submission that *_fast and
*_unlocked look very much the same. I agree and I think further cleanup
will be coming. But I'm focused on getting the final solution for DAX at
the moment.
This patch (of 7):
This patch starts a series which aims to support FOLL_LONGTERM in
get_user_pages_fast(). Some callers who would like to do a longterm (user
controlled pin) of pages with the fast variant of GUP for performance
purposes.
Rather than have a separate get_user_pages_longterm() call, introduce
FOLL_LONGTERM and change the longterm callers to use it.
This patch does not change any functionality. In the short term
"longterm" or user controlled pins are unsafe for Filesystems and FS DAX
in particular has been blocked. However, callers of get_user_pages_fast()
were not "protected".
FOLL_LONGTERM can _only_ be supported with get_user_pages[_fast]() as it
requires vmas to determine if DAX is in use.
NOTE: In merging with the CMA changes we opt to change the
get_user_pages() call in check_and_migrate_cma_pages() to a call of
__get_user_pages_locked() on the newly migrated pages. This makes the
code read better in that we are calling __get_user_pages_locked() on the
pages before and after a potential migration.
As a side affect some of the interfaces are cleaned up but this is not the
primary purpose of the series.
In review[1] it was asked:
<quote>
> This I don't get - if you do lock down long term mappings performance
> of the actual get_user_pages call shouldn't matter to start with.
>
> What do I miss?
A couple of points.
First "longterm" is a relative thing and at this point is probably a
misnomer. This is really flagging a pin which is going to be given to
hardware and can't move. I've thought of a couple of alternative names
but I think we have to settle on if we are going to use FL_LAYOUT or
something else to solve the "longterm" problem. Then I think we can
change the flag to a better name.
Second, It depends on how often you are registering memory. I have spoken
with some RDMA users who consider MR in the performance path... For the
overall application performance. I don't have the numbers as the tests
for HFI1 were done a long time ago. But there was a significant
advantage. Some of which is probably due to the fact that you don't have
to hold mmap_sem.
Finally, architecturally I think it would be good for everyone to use
*_fast. There are patches submitted to the RDMA list which would allow
the use of *_fast (they reworking the use of mmap_sem) and as soon as they
are accepted I'll submit a patch to convert the RDMA core as well. Also
to this point others are looking to use *_fast.
As an asside, Jasons pointed out in my previous submission that *_fast and
*_unlocked look very much the same. I agree and I think further cleanup
will be coming. But I'm focused on getting the final solution for DAX at
the moment.
</quote>
[1] https://lore.kernel.org/lkml/20190220180255.GA12020@iweiny-DESK2.sc.intel.com/T/#md6abad2569f3bf6c1f03686c8097ab6563e94965
[ira.weiny@intel.com: v3]
Link: http://lkml.kernel.org/r/20190328084422.29911-2-ira.weiny@intel.com
Link: http://lkml.kernel.org/r/20190328084422.29911-2-ira.weiny@intel.com
Link: http://lkml.kernel.org/r/20190317183438.2057-2-ira.weiny@intel.com
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Rich Felker <dalias@libc.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Mike Marshall <hubcap@omnibond.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The ring buffer code of XDP sockets is missing a memory barrier on the
consumer side between the load of the data and the write that signals
that it is ok for the producer to put new data into the buffer. On
architectures that does not guarantee that stores are not reordered
with older loads, the producer might put data into the ring before the
consumer had the chance to read it. As IA does guarantee this
ordering, it would only need a compiler barrier here, but there are no
primitives in Linux for this specific case (hinder writes to be ordered
before older reads) so I had to add a smp_mb() here which will
translate into a run-time synch operation on IA.
Added a longish comment in the code explaining what each barrier in
the ring implementation accomplishes and what would happen if we
removed one of them.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When the umem is cleaned up, the task that created it might already be
gone. If the task was gone, the xdp_umem_release function did not free
the pages member of struct xdp_umem.
It turned out that the task lookup was not needed at all; The code was
a left-over when we moved from task accounting to user accounting [1].
This patch fixes the memory leak by removing the task lookup logic
completely.
[1] https://lore.kernel.org/netdev/20180131135356.19134-3-bjorn.topel@gmail.com/
Link: https://lore.kernel.org/netdev/c1cb2ca8-6a14-3980-8672-f3de0bb38dfd@suse.cz/
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Reported-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Passing a non-existing option in the options member of struct
xdp_desc was, incorrectly, silently ignored. This patch addresses
that behavior, and drops any Tx descriptor with non-existing options.
We have examined existing user space code, and to our best knowledge,
no one is relying on the current incorrect behavior. AF_XDP is still
in its infancy, so from our perspective, the risk of breakage is very
low, and addressing this problem now is important.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Passing a non-existing flag in the sxdp_flags member of struct
sockaddr_xdp was, incorrectly, silently ignored. This patch addresses
that behavior, and rejects any non-existing flags.
We have examined existing user space code, and to our best knowledge,
no one is relying on the current incorrect behavior. AF_XDP is still
in its infancy, so from our perspective, the risk of breakage is very
low, and addressing this problem now is important.
Fixes: 965a990984 ("xsk: add support for bind for Rx")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Three conflicts, one of which, for marvell10g.c is non-trivial and
requires some follow-up from Heiner or someone else.
The issue is that Heiner converted the marvell10g driver over to
use the generic c45 code as much as possible.
However, in 'net' a bug fix appeared which makes sure that a new
local mask (MDIO_AN_10GBT_CTRL_ADV_NBT_MASK) with value 0x01e0
is cleared.
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit e2ce367488.
It turns out that the sock destructor xsk_destruct was needed after
all. The cleanup simplification broke the skb transmit cleanup path,
due to that the umem was prematurely destroyed.
The umem cannot be destroyed until all outstanding skbs are freed,
which means that we cannot remove the umem until the sk_destruct has
been called.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit c9b47cc1fa ("xsk: fix bug when trying to use both copy and
zero-copy on one queue id") stores the umem into the netdev._rx
struct. However, the patch incorrectly removed the umem from the
netdev._rx struct when user-space passed "best-effort" mode
(i.e. select the fastest possible option available), and zero-copy
mode was not available. This commit fixes that.
Fixes: c9b47cc1fa ("xsk: fix bug when trying to use both copy and zero-copy on one queue id")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Holding mmap_sem exclusively for a gup() is an overkill. Lets
share the lock and replace the gup call for gup_longterm(), as
it is better suited for the lifetime of the pinning.
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: Bjorn Topel <bjorn.topel@intel.com>
Cc: Magnus Karlsson <magnus.karlsson@intel.com>
CC: netdev@vger.kernel.org
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
All the setup code in AF_XDP is protected by a mutex with the
exception of the mmap code that cannot use it. To make sure that a
process banging on the mmap call at the same time as another process
is setting up the socket, smp_wmb() calls were added in the umem
registration code and the queue creation code, so that the published
structures that xsk_mmap needs would be consistent. However, the
corresponding smp_rmb() calls were not added to the xsk_mmap
code. This patch adds these calls.
Fixes: 37b076933a ("xsk: add missing write- and data-dependency barrier")
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Daniel Borkmann says:
====================
pull-request: bpf-next 2019-01-29
The following pull-request contains BPF updates for your *net-next* tree.
The main changes are:
1) Teach verifier dead code removal, this also allows for optimizing /
removing conditional branches around dead code and to shrink the
resulting image. Code store constrained architectures like nfp would
have hard time doing this at JIT level, from Jakub.
2) Add JMP32 instructions to BPF ISA in order to allow for optimizing
code generation for 32-bit sub-registers. Evaluation shows that this
can result in code reduction of ~5-20% compared to 64 bit-only code
generation. Also add implementation for most JITs, from Jiong.
3) Add support for __int128 types in BTF which is also needed for
vmlinux's BTF conversion to work, from Yonghong.
4) Add a new command to bpftool in order to dump a list of BPF-related
parameters from the system or for a specific network device e.g. in
terms of available prog/map types or helper functions, from Quentin.
5) Add AF_XDP sock_diag interface for querying sockets from user
space which provides information about the RX/TX/fill/completion
rings, umem, memory usage etc, from Björn.
6) Add skb context access for skb_shared_info->gso_segs field, from Eric.
7) Add support for testing flow dissector BPF programs by extending
existing BPF_PROG_TEST_RUN infrastructure, from Stanislav.
8) Split BPF kselftest's test_verifier into various subgroups of tests
in order better deal with merge conflicts in this area, from Jakub.
9) Add support for queue/stack manipulations in bpftool, from Stanislav.
10) Document BTF, from Yonghong.
11) Dump supported ELF section names in libbpf on program load
failure, from Taeung.
12) Silence a false positive compiler warning in verifier's BTF
handling, from Peter.
13) Fix help string in bpftool's feature probing, from Prashant.
14) Remove duplicate includes in BPF kselftests, from Yue.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the sock_diag interface for querying sockets from user
space. Tools like iproute2 ss(8) can use this interface to list open
AF_XDP sockets.
The user-space ABI is defined in linux/xdp_diag.h and includes netlink
request and response structs. The request can query sockets and the
response contains socket information about the rings, umems, inode and
more.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds an id to the umem structure. The id uniquely
identifies a umem instance, and will be exposed to user-space via the
socket monitoring interface.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Track each AF_XDP socket in a per-netns list. This will be used later
by the sock_diag interface for querying sockets from userspace.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Export xdp_get_umem_from_qid for other modules to use.
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
In the xdp_umem_assign_dev() path, the xsk code does not
check if a queue for which umem is to be created exists.
It leads to a situation where umem is not assigned to any
Tx/Rx queue of a netdevice, without notifying the stack
about an error. This affects both XDP_SKB and XDP_DRV
modes - in case of XDP_DRV_ZC, queue index is checked by
the driver.
This patch fixes xsk code, so that in both XDP_SKB and
XDP_DRV mode of AF_XDP, an error is returned when requested
queue index exceedes an existing maximum.
Fixes: c9b47cc1fa ("xsk: fix bug when trying to use both copy and zero-copy on one queue id")
Reported-by: Jakub Spizewski <jakub.spizewski@intel.com>
Signed-off-by: Krzysztof Kazimierczak <krzysztof.kazimierczak@intel.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Prior this commit, when the struct socket object was being released,
the UMEM did not have its reference count decreased. Instead, this was
done in the struct sock sk_destruct function.
There is no reason to keep the UMEM reference around when the socket
is being orphaned, so in this patch the xdp_put_mem is called in the
xsk_release function. This results in that the xsk_destruct function
can be removed!
Note that, it still holds that a struct xsk_sock reference might still
linger in the XSKMAP after the UMEM is released, e.g. if a user does
not clear the XSKMAP prior to closing the process. This sock will be
in a "released" zombie like state, until the XSKMAP is removed.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
net/sched/cls_api.c has overlapping changes to a call to
nlmsg_parse(), one (from 'net') added rtm_tca_policy instead of NULL
to the 5th argument, and another (from 'net-next') added cb->extack
instead of NULL to the 6th argument.
net/ipv4/ipmr_base.c is a case of a bug fix in 'net' being done to
code which moved (to mr_table_dump)) in 'net-next'. Thanks to David
Ahern for the heads up.
Signed-off-by: David S. Miller <davem@davemloft.net>
The XSKMAP update and delete functions called synchronize_net(), which
can sleep. It is not allowed to sleep during an RCU read section.
Instead we need to make sure that the sock sk_destruct (xsk_destruct)
function is asynchronously called after an RCU grace period. Setting
the SOCK_RCU_FREE flag for XDP sockets takes care of this.
Fixes: fbfc504a24 ("bpf: introduce new bpf AF_XDP map type BPF_MAP_TYPE_XSKMAP")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The AF_XDP socket struct can exist in three different, implicit
states: setup, bound and released. Setup is prior the socket has been
bound to a device. Bound is when the socket is active for receive and
send. Released is when the process/userspace side of the socket is
released, but the sock object is still lingering, e.g. when there is a
reference to the socket in an XSKMAP after process termination.
The Rx fast-path code uses the "dev" member of struct xdp_sock to
check whether a socket is bound or relased, and the Tx code uses the
struct xdp_umem "xsk_list" member in conjunction with "dev" to
determine the state of a socket.
However, the transition from bound to released did not tear the socket
down in correct order.
On the Rx side "dev" was cleared after synchronize_net() making the
synchronization useless. On the Tx side, the internal queues were
destroyed prior removing them from the "xsk_list".
This commit corrects the cleanup order, and by doing so
xdp_del_sk_umem() can be simplified and one synchronize_net() can be
removed.
Fixes: 965a990984 ("xsk: add support for bind for Rx")
Fixes: ac98d8aab6 ("xsk: wire upp Tx zero-copy functions")
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
As we now do not allow ethtool to deactivate the queue id we are
running an AF_XDP socket on, we can simplify the implementation of
xdp_clear_umem_at_qid().
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
We already check the RSS indirection table does not use queues which
would be disabled by channel reconfiguration. Make sure user does not
try to disable queues which have a UMEM and zero-copy AF_XDP socket
installed.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously, the xsk code did not record which umem was bound to a
specific queue id. This was not required if all drivers were zero-copy
enabled as this had to be recorded in the driver anyway. So if a user
tried to bind two umems to the same queue, the driver would say
no. But if copy-mode was first enabled and then zero-copy mode (or the
reverse order), we mistakenly enabled both of them on the same umem
leading to buggy behavior. The main culprit for this is that we did
not store the association of umem to queue id in the copy case and
only relied on the driver reporting this. As this relation was not
stored in the driver for copy mode (it does not rely on the AF_XDP
NDOs), this obviously could not work.
This patch fixes the problem by always recording the umem to queue id
relationship in the netdev_queue and netdev_rx_queue structs. This way
we always know what kind of umem has been bound to a queue id and can
act appropriately at bind time.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
XSK UMEM is strongly single producer single consumer so reuse of
frames is challenging. Add a simple "stash" of FILL packets to
reuse for drivers to optionally make use of. This is useful
when driver has to free (ndo_stop) or resize a ring with an active
AF_XDP ZC socket.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This commit gets rid of the structure xdp_umem_props. It was there to
be able to break a dependency at one point, but this is no longer
needed. The values in the struct are instead stored directly in the
xdp_umem structure. This simplifies the xsk code as well as af_xdp
zero-copy drivers and as a bonus gets rid of one internal header file.
The i40e driver is also adapted to the new interface in this commit.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Since xdp_umem_query() was added one assignment of bpf.command was
missed from cleanup. Removing the assignment statement.
Fixes: 84c6b86875 ("xsk: don't allow umem replace at stack level")
Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously, the AF_XDP (XDP_DRV/XDP_SKB copy-mode) ingress logic did
not include XDP meta data in the data buffers copied out to the user
application.
In this commit, we check if meta data is available, and if so, it is
prepended to the frame.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Move the xdp_umem_get_{data,dma} functions to include/net/xdp_sock.h,
so that the upcoming zero-copy implementation in the Ethernet drivers
can utilize them.
Also, supply some dummy function implementations for
CONFIG_XDP_SOCKETS=n configs.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
s/ENOTSUPP/EOPNOTSUPP/ in function umem_assign_dev().
This function's return value is directly returned by xsk_bind().
EOPNOTSUPP is bind()'s possible return value.
Fixes: f734607e81 ("xsk: refactor xdp_umem_assign_dev()")
Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The BTF conflicts were simple overlapping changes.
The virtio_net conflict was an overlap of a fix of statistics counter,
happening alongisde a move over to a bonafide statistics structure
rather than counting value on the stack.
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently drivers have to check if they already have a umem
installed for a given queue and return an error if so. Make
better use of XDP_QUERY_XSK_UMEM and move this functionality
to the core.
We need to keep rtnl across the calls now.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Björn Töpel <bjorn.topel@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Return early and only take the ref on dev once there is no possibility
of failing.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Björn Töpel <bjorn.topel@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
xdp_return_buff() is used when frame has been successfully
handled (transmitted) or if an error occurred during delayed
processing and there is no way to report it back to
xdp_do_redirect().
In case of __xsk_rcv_zc() error is propagated all the way
back to the driver, so there is no need to call
xdp_return_buff(). Driver will recycle the frame anyway
after seeing that error happened.
Fixes: 173d3adb6f ("xsk: add zero-copy support for Rx")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Polling for the ingress queues relies on reading the producer/consumer
pointers of the Rx queue.
Prior this commit, a cached consumer pointer could be used, instead of
the actual consumer pointer and therefore report POLLIN prematurely.
This patch makes sure that the non-cached consumer pointer is used
instead.
Reported-by: Qi Zhang <qi.z.zhang@intel.com>
Tested-by: Qi Zhang <qi.z.zhang@intel.com>
Fixes: c497176cb2 ("xsk: add Rx receive functions and poll support")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch stops returning EMSGSIZE from sendmsg in copy mode when the
size of the packet is larger than the MTU. Just send it to the device
so that it will drop it as in zero-copy mode. This makes the error
reporting consistent between copy mode and zero-copy mode.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch makes sure ENOBUFS is always returned from sendmsg if there
is no TX queue configured. This was not the case for zero-copy
mode. With this patch this error reporting is consistent between copy
mode and zero-copy mode.
Fixes: ac98d8aab6 ("xsk: wire upp Tx zero-copy functions")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch stops returning EAGAIN in TX copy mode when the completion
queue is full as zero-copy does not do this. Instead this situation
can be detected by comparing the head and tail pointers of the
completion queue in both modes. In any case, EAGAIN was not the
correct error code here since no amount of calling sendmsg will solve
the problem. Only consuming one or more messages on the completion
queue will fix this.
With this patch, the error reporting becomes consistent between copy
mode and zero-copy mode.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch removes the ENXIO return code from TX copy-mode when
someone has forcefully changed the number of queues on the device so
that the queue bound to the socket is no longer available. Just
silently stop sending anything as in zero-copy mode so the error
reporting gets consistent between the two modes.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
There is a potential race in the TX completion code for the SKB
case. One process enters the sendmsg code of an AF_XDP socket in order
to send a frame. The execution eventually trickles down to the driver
that is told to send the packet. However, it decides to drop the
packet due to some error condition (e.g., rings full) and frees the
SKB. This will trigger the SKB destructor and a completion will be
sent to the AF_XDP user space through its
single-producer/single-consumer queues.
At the same time a TX interrupt has fired on another core and it
dispatches the TX completion code in the driver. It does its HW
specific things and ends up freeing the SKB associated with the
transmitted packet. This will trigger the SKB destructor and a
completion will be sent to the AF_XDP user space through its
single-producer/single-consumer queues. With a pseudo call stack, it
would look like this:
Core 1:
sendmsg() being called in the application
netdev_start_xmit()
Driver entered through ndo_start_xmit
Driver decides to free the SKB for some reason (e.g., rings full)
Destructor of SKB called
xskq_produce_addr() is called to signal completion to user space
Core 2:
TX completion irq
NAPI loop
Driver irq handler for TX completions
Frees the SKB
Destructor of SKB called
xskq_produce_addr() is called to signal completion to user space
We now have a violation of the single-producer/single-consumer
principle for our queues as there are two threads trying to produce at
the same time on the same queue.
Fixed by introducing a spin_lock in the destructor. In regards to the
performance, I get around 1.74 Mpps for txonly before and after the
introduction of the spinlock. There is of course some impact due to
the spin lock but it is in the less significant digits that are too
noisy for me to measure. But let us say that the version without the
spin lock got 1.745 Mpps in the best case and the version with 1.735
Mpps in the worst case, then that would mean a maximum drop in
performance of 0.5%.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Fixed a bug in which a frame could be completed more than once
when an error was returned from dev_direct_xmit(). The code
erroneously retried sending the message leading to multiple
calls to the SKB destructor and therefore multiple completions
of the same buffer to user space.
The error code in this case has been changed from EAGAIN to EBUSY
in order to tell user space that the sending of the packet failed
and the buffer has been return to user space through the completion
queue.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Reported-by: Pavel Odintsov <pavel@fastnetmon.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The code in xskq_produce_addr erroneously checked if there
was up to LAZY_UPDATE_THRESHOLD amount of space in the completion
queue. It only needs to check if there is one slot left in the
queue. This bug could under some circumstances lead to a WARN_ON_ONCE
being triggered and the completion message to user space being lost.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Reported-by: Pavel Odintsov <pavel@fastnetmon.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The poll() changes were not well thought out, and completely
unexplained. They also caused a huge performance regression, because
"->poll()" was no longer a trivial file operation that just called down
to the underlying file operations, but instead did at least two indirect
calls.
Indirect calls are sadly slow now with the Spectre mitigation, but the
performance problem could at least be largely mitigated by changing the
"->get_poll_head()" operation to just have a per-file-descriptor pointer
to the poll head instead. That gets rid of one of the new indirections.
But that doesn't fix the new complexity that is completely unwarranted
for the regular case. The (undocumented) reason for the poll() changes
was some alleged AIO poll race fixing, but we don't make the common case
slower and more complex for some uncommon special case, so this all
really needs way more explanations and most likely a fundamental
redesign.
[ This revert is a revert of about 30 different commits, not reverted
individually because that would just be unnecessarily messy - Linus ]
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit 173d3adb6f ("xsk: add zero-copy support for Rx") introduced a
regression on the XDP_SKB receive path, when the queue id checks were
removed. Now, they are back again.
Fixes: 173d3adb6f ("xsk: add zero-copy support for Rx")
Reported-by: Qi Zhang <qi.z.zhang@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
syzkaller reported a warning from xdp_umem_pin_pages():
WARNING: CPU: 1 PID: 4537 at mm/slab_common.c:996 kmalloc_slab+0x56/0x70 mm/slab_common.c:996
...
__do_kmalloc mm/slab.c:3713 [inline]
__kmalloc+0x25/0x760 mm/slab.c:3727
kmalloc_array include/linux/slab.h:634 [inline]
kcalloc include/linux/slab.h:645 [inline]
xdp_umem_pin_pages net/xdp/xdp_umem.c:205 [inline]
xdp_umem_reg net/xdp/xdp_umem.c:318 [inline]
xdp_umem_create+0x5c9/0x10f0 net/xdp/xdp_umem.c:349
xsk_setsockopt+0x443/0x550 net/xdp/xsk.c:531
__sys_setsockopt+0x1bd/0x390 net/socket.c:1935
__do_sys_setsockopt net/socket.c:1946 [inline]
__se_sys_setsockopt net/socket.c:1943 [inline]
__x64_sys_setsockopt+0xbe/0x150 net/socket.c:1943
do_syscall_64+0x1b1/0x800 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x49/0xbe
This is a warning about attempting to allocate more than
KMALLOC_MAX_SIZE memory. The request originates from userspace, and if
the request is too big, the kernel is free to deny its allocation. In
this patch, the failed allocation attempt is silenced with
__GFP_NOWARN.
Fixes: c0c77d8fb7 ("xsk: add user memory registration support sockopt")
Reported-by: syzbot+4abadc5d69117b346506@syzkaller.appspotmail.com
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
syzkaller was able to trigger the following panic for AF_XDP:
BUG: KASAN: null-ptr-deref in atomic64_sub include/asm-generic/atomic-instrumented.h:144 [inline]
BUG: KASAN: null-ptr-deref in atomic_long_sub include/asm-generic/atomic-long.h:199 [inline]
BUG: KASAN: null-ptr-deref in xdp_umem_unaccount_pages.isra.4+0x3d/0x80 net/xdp/xdp_umem.c:135
Write of size 8 at addr 0000000000000060 by task syz-executor246/4527
CPU: 1 PID: 4527 Comm: syz-executor246 Not tainted 4.17.0+ #89
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0x1b9/0x294 lib/dump_stack.c:113
kasan_report_error mm/kasan/report.c:352 [inline]
kasan_report.cold.7+0x6d/0x2fe mm/kasan/report.c:412
check_memory_region_inline mm/kasan/kasan.c:260 [inline]
check_memory_region+0x13e/0x1b0 mm/kasan/kasan.c:267
kasan_check_write+0x14/0x20 mm/kasan/kasan.c:278
atomic64_sub include/asm-generic/atomic-instrumented.h:144 [inline]
atomic_long_sub include/asm-generic/atomic-long.h:199 [inline]
xdp_umem_unaccount_pages.isra.4+0x3d/0x80 net/xdp/xdp_umem.c:135
xdp_umem_reg net/xdp/xdp_umem.c:334 [inline]
xdp_umem_create+0xd6c/0x10f0 net/xdp/xdp_umem.c:349
xsk_setsockopt+0x443/0x550 net/xdp/xsk.c:531
__sys_setsockopt+0x1bd/0x390 net/socket.c:1935
__do_sys_setsockopt net/socket.c:1946 [inline]
__se_sys_setsockopt net/socket.c:1943 [inline]
__x64_sys_setsockopt+0xbe/0x150 net/socket.c:1943
do_syscall_64+0x1b1/0x800 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x49/0xbe
In xdp_umem_reg() the call to xdp_umem_account_pages() passed
with CAP_IPC_LOCK where we didn't need to end up charging rlimit
on memlock for the current user and therefore umem->user continues
to be NULL. Later on through fault injection syzkaller triggered
a failure in either umem->pgs or umem->pages allocation such that
we bail out and undo accounting in xdp_umem_unaccount_pages()
where we eventually hit the panic since it tries to deref the
umem->user.
The code is pretty close to mm_account_pinned_pages() and
mm_unaccount_pinned_pages() pair and potentially could reuse
it even in a later cleanup, and it appears that the initial
commit c0c77d8fb7 ("xsk: add user memory registration support
sockopt") got this right while later follow-up introduced the
bug via a49049ea25 ("xsk: simplified umem setup").
Fixes: a49049ea25 ("xsk: simplified umem setup")
Reported-by: syzbot+979217770b09ebf5c407@syzkaller.appspotmail.com
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
With gcc-4.1.2 on 32-bit:
net/xdp/xsk.c:663: warning: integer constant is too large for ‘long’ type
net/xdp/xsk.c:665: warning: integer constant is too large for ‘long’ type
Add the missing "ULL" suffixes to the large XDP_UMEM_PGOFF_*_RING values
to fix this.
net/xdp/xsk.c:663: warning: comparison is always false due to limited range of data type
net/xdp/xsk.c:665: warning: comparison is always false due to limited range of data type
"unsigned long" is 32-bit on 32-bit systems, hence the offset is
truncated, and can never be equal to any of the XDP_UMEM_PGOFF_*_RING
values. Use loff_t (and the required cast) to fix this.
Fixes: 423f38329d ("xsk: add umem fill queue support and mmap")
Fixes: fe2308328c ("xsk: add umem completion queue support and mmap")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Pull networking updates from David Miller:
1) Add Maglev hashing scheduler to IPVS, from Inju Song.
2) Lots of new TC subsystem tests from Roman Mashak.
3) Add TCP zero copy receive and fix delayed acks and autotuning with
SO_RCVLOWAT, from Eric Dumazet.
4) Add XDP_REDIRECT support to mlx5 driver, from Jesper Dangaard
Brouer.
5) Add ttl inherit support to vxlan, from Hangbin Liu.
6) Properly separate ipv6 routes into their logically independant
components. fib6_info for the routing table, and fib6_nh for sets of
nexthops, which thus can be shared. From David Ahern.
7) Add bpf_xdp_adjust_tail helper, which can be used to generate ICMP
messages from XDP programs. From Nikita V. Shirokov.
8) Lots of long overdue cleanups to the r8169 driver, from Heiner
Kallweit.
9) Add BTF ("BPF Type Format"), from Martin KaFai Lau.
10) Add traffic condition monitoring to iwlwifi, from Luca Coelho.
11) Plumb extack down into fib_rules, from Roopa Prabhu.
12) Add Flower classifier offload support to igb, from Vinicius Costa
Gomes.
13) Add UDP GSO support, from Willem de Bruijn.
14) Add documentation for eBPF helpers, from Quentin Monnet.
15) Add TLS tx offload to mlx5, from Ilya Lesokhin.
16) Allow applications to be given the number of bytes available to read
on a socket via a control message returned from recvmsg(), from
Soheil Hassas Yeganeh.
17) Add x86_32 eBPF JIT compiler, from Wang YanQing.
18) Add AF_XDP sockets, with zerocopy support infrastructure as well.
From Björn Töpel.
19) Remove indirect load support from all of the BPF JITs and handle
these operations in the verifier by translating them into native BPF
instead. From Daniel Borkmann.
20) Add GRO support to ipv6 gre tunnels, from Eran Ben Elisha.
21) Allow XDP programs to do lookups in the main kernel routing tables
for forwarding. From David Ahern.
22) Allow drivers to store hardware state into an ELF section of kernel
dump vmcore files, and use it in cxgb4. From Rahul Lakkireddy.
23) Various RACK and loss detection improvements in TCP, from Yuchung
Cheng.
24) Add TCP SACK compression, from Eric Dumazet.
25) Add User Mode Helper support and basic bpfilter infrastructure, from
Alexei Starovoitov.
26) Support ports and protocol values in RTM_GETROUTE, from Roopa
Prabhu.
27) Support bulking in ->ndo_xdp_xmit() API, from Jesper Dangaard
Brouer.
28) Add lots of forwarding selftests, from Petr Machata.
29) Add generic network device failover driver, from Sridhar Samudrala.
* ra.kernel.org:/pub/scm/linux/kernel/git/davem/net-next: (1959 commits)
strparser: Add __strp_unpause and use it in ktls.
rxrpc: Fix terminal retransmission connection ID to include the channel
net: hns3: Optimize PF CMDQ interrupt switching process
net: hns3: Fix for VF mailbox receiving unknown message
net: hns3: Fix for VF mailbox cannot receiving PF response
bnx2x: use the right constant
Revert "net: sched: cls: Fix offloading when ingress dev is vxlan"
net: dsa: b53: Fix for brcm tag issue in Cygnus SoC
enic: fix UDP rss bits
netdev-FAQ: clarify DaveM's position for stable backports
rtnetlink: validate attributes in do_setlink()
mlxsw: Add extack messages for port_{un, }split failures
netdevsim: Add extack error message for devlink reload
devlink: Add extack to reload and port_{un, }split operations
net: metrics: add proper netlink validation
ipmr: fix error path when ipmr_new_table fails
ip6mr: only set ip6mr_table from setsockopt when ip6mr_new_table succeeds
net: hns3: remove unused hclgevf_cfg_func_mta_filter
netfilter: provide udp*_lib_lookup for nf_tproxy
qed*: Utilize FW 8.37.2.0
...
Here we add the functionality required to support zero-copy Tx, and
also exposes various zero-copy related functions for the netdevs.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Extend the xsk_rcv to support the new MEM_TYPE_ZERO_COPY memory, and
wireup ndo_bpf call in bind.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The xdp_umem_page holds the address for a page. Trade memory for
faster lookup. Later, we'll add DMA address here as well.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Moved struct xdp_umem to xdp_sock.h, in order to prepare for zero-copy
support.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently, AF_XDP only supports a fixed frame-size memory scheme where
each frame is referenced via an index (idx). A user passes the frame
index to the kernel, and the kernel acts upon the data. Some NICs,
however, do not have a fixed frame-size model, instead they have a
model where a memory window is passed to the hardware and multiple
frames are filled into that window (referred to as the "type-writer"
model).
By changing the descriptor format from the current frame index
addressing scheme, AF_XDP can in the future be extended to support
these kinds of NICs.
In the index-based model, an idx refers to a frame of size
frame_size. Addressing a frame in the UMEM is done by offseting the
UMEM starting address by a global offset, idx * frame_size + offset.
Communicating via the fill- and completion-rings are done by means of
idx.
In this commit, the idx is removed in favor of an address (addr),
which is a relative address ranging over the UMEM. To convert an
idx-based address to the new addr is simply: addr = idx * frame_size +
offset.
We also stop referring to the UMEM "frame" as a frame. Instead it is
simply called a chunk.
To transfer ownership of a chunk to the kernel, the addr of the chunk
is passed in the fill-ring. Note, that the kernel will mask addr to
make it chunk aligned, so there is no need for userspace to do
that. E.g., for a chunk size of 2k, passing an addr of 2048, 2050 or
3000 to the fill-ring will refer to the same chunk.
On the completion-ring, the addr will match that of the Tx descriptor,
passed to the kernel.
Changing the descriptor format to use chunks/addr will allow for
future changes to move to a type-writer based model, where multiple
frames can reside in one chunk. In this model passing one single chunk
into the fill-ring, would potentially result in multiple Rx
descriptors.
This commit changes the uapi of AF_XDP sockets, and updates the
documentation.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously, rx_dropped could be updated incorrectly, e.g. if the XDP
program redirected the frame to a socket bound to a different queue
than where the XDP program was executing.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously the fill queue descriptor was not copied to kernel space
prior validating it, making it possible for userland to change the
descriptor post-kernel-validation.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
As suggested by Daniel Borkmann, the umem setup code was a too
defensive and complex. Here, we reduce the number of checks. Also, the
memory pinning is now folded into the umem creation, and we do correct
locking.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Here, we add a missing write-barrier, and use READ_ONCE for the
data-dependency barrier.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
