On Sat, Apr 28, 2018 at 08:41:28PM -0700, John Fastabend wrote:
> Sockmap is currently backed by an array and enforces keys to be
> four bytes. This works well for many use cases and was originally
> modeled after devmap which also uses four bytes keys. However,
> this has become limiting in larger use cases where a hash would
> be more appropriate. For example users may want to use the 5-tuple
> of the socket as the lookup key.
>
> To support this add hash support.
>
> Signed-off-by: John Fastabend <john.fastab...@gmail.com>
...
> enum bpf_prog_type {
> @@ -1835,7 +1836,10 @@ struct bpf_stack_build_id {
> FN(msg_pull_data), \
> FN(bind), \
> FN(xdp_adjust_tail), \
> - FN(skb_get_xfrm_state),
> + FN(skb_get_xfrm_state), \
> + FN(sock_hash_update), \
> + FN(msg_redirect_hash), \
> + FN(sk_redirect_hash),
Documentation for new helpers is missing. Please add it in this commit.
Also running test_sockmap in the latest bpf-next I see:
[TEST 81]: (10, 1024, 256, sendpage, pass,apply,): [ 14.227128] WARNING: CPU:
1 PID: 202 at ../net/core/stream.c:206 sk_stream_kill_queues+0x3ca/0x540
[ 14.228209] Modules linked in:
[ 14.228583] CPU: 1 PID: 202 Comm: test_sockmap Not tainted
4.17.0-rc2-00438-gfcf85729d8ef #941
[ 14.229595] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.9.3-1.el7.centos 04/01/2014
[ 14.230649] RIP: 0010:sk_stream_kill_queues+0x3ca/0x540
[ 14.231251] RSP: 0018:ffff880111717d00 EFLAGS: 00010202
[ 14.231855] RAX: 0000000000000000 RBX: ffff8801127f8978 RCX: 1ffff100224ff138
[ 14.232664] RDX: 0000000000000000 RSI: 0000000000000f01 RDI: ffff8801127f89c0
[ 14.233521] RBP: ffff8801127f8948 R08: ffffed00224ff122 R09: ffffed00224ff121
[ 14.234342] R10: ffff8801127f890b R11: ffffed00224ff122 R12: ffff8801127f8998
[ 14.235169] R13: ffff8801127f88a8 R14: ffff8801127f8880 R15: ffff8801127f8930
[ 14.235987] FS: 00007f3d46898700(0000) GS:ffff88011b080000(0000)
knlGS:0000000000000000
[ 14.236946] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 14.237607] CR2: 00007f3d45e3caf0 CR3: 0000000112da0001 CR4: 00000000000606e0
[ 14.238433] Call Trace:
[ 14.238741] ? tcp_v4_destroy_sock+0x273/0x4d0
[ 14.239259] inet_csk_destroy_sock+0x157/0x390
[ 14.239815] tcp_close+0x683/0xf60
[ 14.240214] inet_release+0xd6/0x1b0
[ 14.240636] sock_release+0x7e/0x190
[ 14.241090] sock_close+0xe/0x20
[ 14.241593] __fput+0x22f/0x720
[ 14.241981] task_work_run+0x10b/0x190
[ 14.242425] exit_to_usermode_loop+0xd2/0xf0
[ 14.242963] do_syscall_64+0x1d5/0x270
[ 14.243410] ? page_fault+0x8/0x30
[ 14.243830] entry_SYSCALL_64_after_hwframe+0x44/0xa9
and
[TEST 84]: (1, 1, 1, sendmsg, pass,cork,): [ 14.317100] WARNING: CPU: 1 PID:
66 at ../net/ipv4/af_inet.c:156 inet_sock_destruct+0x570/0x7b0
[ 14.318147] Modules linked in:
[ 14.318522] CPU: 1 PID: 66 Comm: kworker/1:1 Tainted: G W
4.17.0-rc2-00438-gfcf85729d8ef #941
[ 14.319674] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.9.3-1.el7.centos 04/01/2014
[ 14.320764] Workqueue: events smap_gc_work
[ 14.321262] RIP: 0010:inet_sock_destruct+0x570/0x7b0
[ 14.321829] RSP: 0018:ffff880118617d08 EFLAGS: 00010206
[ 14.322415] RAX: 0000000000000ffc RBX: ffff8801127f9980 RCX: ffffffff82c7ac64
[ 14.323221] RDX: 1ffff100224ff34f RSI: 0000000000000004 RDI: ffff8801127f9a78
[ 14.324054] RBP: ffff8801127f9ac4 R08: ffffed00224ff359 R09: ffffed00224ff358
[ 14.324860] R10: ffff8801127f9ac7 R11: ffffed00224ff359 R12: ffff880111107688
[ 14.325661] R13: dffffc0000000000 R14: ffff880111107688 R15: dead000000000100
[ 14.326503] FS: 0000000000000000(0000) GS:ffff88011b080000(0000)
knlGS:0000000000000000
[ 14.327413] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 14.328091] CR2: 00007f3d45e3caf0 CR3: 0000000003a0e004 CR4: 00000000000606e0
[ 14.328936] Call Trace:
[ 14.329225] __sk_destruct+0x45/0x460
[ 14.329646] smap_gc_work+0x69f/0x950
[ 14.330102] process_one_work+0x7be/0x1230
[ 14.330600] worker_thread+0xd9/0x1080
[ 14.331059] ? rescuer_thread+0xd70/0xd70
[ 14.331543] kthread+0x29e/0x390
[ 14.331945] ? kthread_create_worker_on_cpu+0xb0/0xb0
[ 14.332555] ret_from_fork+0x1f/0x30
Please send a fix for this issue first.
I'd like to see test_sockmap being clean before adding hash support to sockmap.
