Age | Commit message (Collapse) | Author |
|
The commit 4b638f13bab4 ("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: 4b638f13bab4 ("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
|
|
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
|
|
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 c593642c8be0 ("treewide: Use sizeof_field() macro")
which gets in the way with b590cb5f802d ("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),
>>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c
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),
>>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c
Just take the one with offsetofend() /and/ gso_segs. Latter is correct due to
850a88cc4096 ("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);
>>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c
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)
>>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c
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 01f52e16b868 ("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
|
|
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
|
|
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
|
|
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: 77cd0d7b3f25 ("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
|
|
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
|
|
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: 77cd0d7b3f25 ("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
|
|
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>
|
|
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: 77cd0d7b3f25 ("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: 423f38329d26 ("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: 37b076933a8e ("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>
|
|
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>
|
|
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: 455302d1c9ae ("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: 35fcde7f8deb ("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>
|
|
Two cases of overlapping changes, nothing fancy.
Signed-off-by: David S. Miller <davem@davemloft.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: c497176cb2e4 ("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: 965a99098443 ("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>
|
|
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>
|
|
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: 965a99098443 ("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 e2ce3674883ecba2605370404208c9d4a07ae1c3.
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>
|
|
Two easily resolvable overlapping change conflicts, one in
TCP and one in the eBPF verifier.
Signed-off-by: David S. Miller <davem@davemloft.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: 37b076933a8e3 ("xsk: add missing write- and data-dependency barrier")
Fixes: c0c77d8fb787c ("xsk: add user memory registration support sockopt")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
|
|
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>
|
|
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>
|
|
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: fbfc504a24f5 ("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: 965a99098443 ("xsk: add support for bind for Rx")
Fixes: ac98d8aab61b ("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>
|
|
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>
|
|
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>
|
|
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>
|
|
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: 173d3adb6f43 ("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>
|
|
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: 35fcde7f8deb ("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: ac98d8aab61b ("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: 35fcde7f8deb ("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: 35fcde7f8deb ("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: 35fcde7f8deb ("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: 35fcde7f8deb ("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>
|