[ofa-general] [PATCH] sdp: fix some warning and bugs in porting to ofed 1.5
Amir Vadai
amirv at mellanox.co.il
Thu Jul 16 06:05:31 PDT 2009
Maarten Hi,
I checked it in my setup using kernel 2.6.30 and I got same results as in ofed-1.4.
Since SDP in 1.5 is going to have lots of changes in the data path it might not be very stable those days.
It should be stabilized a lot more around mid August.
Anyway the exact version I used could be taken from:
url: git://git.openfabrics.org/~amirv/ofed_1_5.git
branch: zcopy
There should not be differences between this and ofed_kernel_1_5 branch (which I guess you used).
For some info about the tests I did and if you have something to add - please view and update bugzilla:
https://bugs.openfabrics.org/show_bug.cgi?id=1682
- Amir
On 07/15/2009 03:08 PM, Maarten van Malland wrote:
> Hi Amir,
>
> We tried your patch and unfortunately it didn't help much. Actually, it made
> it worse. Here is the dmesg output when we ran nttcp over sdp:
>
> inftsttwin03 ~ # dmesg
> sdp_reset_sk:447 sdp_sock( 4453:0 5038:51621): setting state to error
> BUG: unable to handle kernel paging request at 0000000000001398
> IP: [<ffffffffa0104ff0>] sdp_post_recv+0x299/0x66b [ib_sdp]
> PGD 11a961067 PUD 12dcac067 PMD 0
> Oops: 0002 [#1] PREEMPT SMP
> last sysfs file: /sys/devices/system/cpu/cpu7/cache/index2/shared_cpu_map
> CPU 5
> Modules linked in: ib_sdp w83793 hwmon_vid rdma_ucm rdma_cm iw_cm ib_addr
> mlx4_ib ib_ipoib ib_cm ib_sa ib_
> uverbs ib_umad ib_mthca ib_mad i2c_i801 ib_core mlx4_core e1000e i2c_core
> i5k_amb hwmon [last unloaded: ib
> _sdp]
> Pid: 12072, comm: nttcp Not tainted 2.6.30.1 #1 X7DWT
> RIP: 0010:[<ffffffffa0104ff0>] [<ffffffffa0104ff0>]
> sdp_post_recv+0x299/0x66b [ib_sdp]
> RSP: 0018:ffff88011b4d1ab8 EFLAGS: 00010202
> RAX: 0000000000001398 RBX: 0000000000000008 RCX: ffffe20003df9d40
> RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffffff80277789
> RBP: ffff88011b4d1bf8 R08: 0000000000000000 R09: ffff880000015e00
> R10: 0000000000000000 R11: ffffe20003df9d40 R12: ffff88011099f500
> R13: ffff880110474080 R14: 00000000000000d2 R15: 0000000000000000
> FS: 00007f337b97c6f0(0000) GS:ffff880028186000(0000) knlGS:0000000000000000
> CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> CR2: 0000000000001398 CR3: 000000012d8bb000 CR4: 00000000000406e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> Process nttcp (pid: 12072, threadinfo ffff88011b4d0000, task
> ffff88011b49e640)
> Stack:
> ffff88011b4d1af8 ffffffff802308a8 00000000ffffffff ffff88012d4b1000
> 0000000000001398 ffff88012d965800 0000000000000139 0000000000011a80
> ffff88011b4d1b98 ffffffff80553141 000000011b4c8000 0000000000000001
> Call Trace:
> [<ffffffff802308a8>] ? finish_task_switch+0xb5/0xc4
> [<ffffffff80553141>] ? thread_return+0x4e/0x9d
> [<ffffffff8023c49e>] ? local_bh_disable+0xd/0xf
> [<ffffffff80554ffa>] ? _spin_lock_bh+0x11/0x34
> [<ffffffff80554f99>] ? _spin_lock_irqsave+0x1d/0x39
> [<ffffffff805579d2>] ? sub_preempt_count+0x9e/0xbe
> [<ffffffffa01060e3>] sdp_do_posts+0xa71/0xa91 [ib_sdp]
> [<ffffffff804c7700>] ? sk_wait_data+0xb8/0xcd
> [<ffffffff8024b10f>] ? autoremove_wake_function+0x0/0x38
> [<ffffffffa00ff006>] sdp_recvmsg+0x360/0x5dc [ib_sdp]
> [<ffffffff804c63ed>] sock_common_recvmsg+0x32/0x47
> [<ffffffff804c3fdb>] sock_aio_read+0xe8/0xfc
> [<ffffffff8023c703>] ? irq_exit+0x4f/0x51
> [<ffffffff802a2334>] ? do_sync_read+0x62/0x126
> [<ffffffff802a23b4>] do_sync_read+0xe2/0x126
> [<ffffffff80554d46>] ? trace_hardirqs_on_thunk+0x3a/0x3c
> [<ffffffff8020b8ad>] ? restore_args+0x0/0x30
> [<ffffffff8024b10f>] ? autoremove_wake_function+0x0/0x38
> [<ffffffff805579d2>] ? sub_preempt_count+0x9e/0xbe
> [<ffffffff80555295>] ? _spin_unlock+0x10/0x29
> [<ffffffff802a2af3>] vfs_read+0xbe/0x134
> [<ffffffff802a2e5d>] sys_read+0x47/0x70
> [<ffffffff8020aeeb>] system_call_fastpath+0x16/0x1b
> Code: b8 00 00 00 41 3b 9d 6c 05 00 00 7c 83 8b 85 f0 fe ff ff 49 8b 95 30
> 04 00 00 83 e0 3f 48 6b c0 58 4
> 8 01 d0 48 89 85 e0 fe ff ff <4c> 89 20 4d 8b bd d0 04 00 00 49 8b 87 98 02
> 00 00 48 85 c0 74
> RIP [<ffffffffa0104ff0>] sdp_post_recv+0x299/0x66b [ib_sdp]
> RSP <ffff88011b4d1ab8>
> CR2: 0000000000001398
> ---[ end trace f3d85b6359a569b5 ]---
>
> Perhaps you can take a second look at it,
>
> Best regards,
>
> Maarten
>
>
> -----Oorspronkelijk bericht-----
> Van: general-bounces at lists.openfabrics.org
> [mailto:general-bounces at lists.openfabrics.org] Namens Amir Vadai
> Verzonden: woensdag 15 juli 2009 12:28
> Aan: general at lists.openfabrics.org
> CC: Amir Vadai
> Onderwerp: [ofa-general] [PATCH] sdp: fix some warning and bugs in porting
> to ofed 1.5
>
> Signed-off-by: Amir Vadai <amirv at mellanox.co.il>
> ---
> drivers/infiniband/ulp/sdp/sdp.h | 9 +++++++--
> drivers/infiniband/ulp/sdp/sdp_cma.c | 2 +-
> drivers/infiniband/ulp/sdp/sdp_main.c | 17 ++++++++---------
> drivers/infiniband/ulp/sdp/sdp_rx.c | 14 ++++----------
> 4 files changed, 20 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/infiniband/ulp/sdp/sdp.h
> b/drivers/infiniband/ulp/sdp/sdp.h
> index d03d6af..ec8dbe1 100644
> --- a/drivers/infiniband/ulp/sdp/sdp.h
> +++ b/drivers/infiniband/ulp/sdp/sdp.h
> @@ -11,12 +11,15 @@
> #define SDPSTATS_ON
> /* #define SDP_PROFILING */
>
> -#define _sdp_printk(func, line, level, sk, format, arg...) \
> +#define _sdp_printk(func, line, level, sk, format, arg...) do {
> \
> + preempt_disable(); \
> printk(level "%s:%d sdp_sock(%5d:%d %d:%d): " format, \
> func, line, \
> current->pid, smp_processor_id(), \
> (sk) ? inet_sk(sk)->num : -1, \
> - (sk) ? ntohs(inet_sk(sk)->dport) : -1, ## arg)
> + (sk) ? ntohs(inet_sk(sk)->dport) : -1, ## arg); \
> + preempt_enable(); \
> +} while (0)
> #define sdp_printk(level, sk, format, arg...) \
> _sdp_printk(__func__, __LINE__, level, sk, format, ## arg)
> #define sdp_warn(sk, format, arg...) \
> @@ -71,6 +74,7 @@ static inline unsigned long long current_nsec(void)
> #define sdp_prf1(sk, s, format, arg...) ({ \
> struct sdpprf_log *l = \
> &sdpprf_log[sdpprf_log_count++ & (SDPPRF_LOG_SIZE - 1)]; \
> + preempt_disable(); \
> l->idx = sdpprf_log_count - 1; \
> l->pid = current->pid; \
> l->sk_num = (sk) ? inet_sk(sk)->num : -1; \
> @@ -81,6 +85,7 @@ static inline unsigned long long current_nsec(void)
> l->time = current_nsec(); \
> l->func = __func__; \
> l->line = __LINE__; \
> + preempt_enable(); \
> 1; \
> })
> #define sdp_prf(sk, s, format, arg...)
> diff --git a/drivers/infiniband/ulp/sdp/sdp_cma.c
> b/drivers/infiniband/ulp/sdp/sdp_cma.c
> index 2354fff..f8a7303 100644
> --- a/drivers/infiniband/ulp/sdp/sdp_cma.c
> +++ b/drivers/infiniband/ulp/sdp/sdp_cma.c
> @@ -308,7 +308,7 @@ int sdp_cma_handler(struct rdma_cm_id *id, struct
> rdma_cm_event *event)
> -EINVAL : 0;
> }
>
> - lock_sock(sk);
> + lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
> sdp_dbg(sk, "%s event %d id %p\n", __func__, event->event, id);
> if (!sdp_sk(sk)->id) {
> sdp_dbg(sk, "socket is being torn down\n");
> diff --git a/drivers/infiniband/ulp/sdp/sdp_main.c
> b/drivers/infiniband/ulp/sdp/sdp_main.c
> index 48e43dc..ab49846 100644
> --- a/drivers/infiniband/ulp/sdp/sdp_main.c
> +++ b/drivers/infiniband/ulp/sdp/sdp_main.c
> @@ -181,7 +181,6 @@ static int sdp_get_port(struct sock *sk, unsigned short
> snum)
> static void sdp_destroy_qp(struct sdp_sock *ssk)
> {
> struct ib_pd *pd = NULL;
> - unsigned long flags;
>
>
> sdp_dbg(&ssk->isk.sk, "destroying qp\n");
> @@ -189,7 +188,6 @@ static void sdp_destroy_qp(struct sdp_sock *ssk)
>
> del_timer(&ssk->tx_ring.timer);
>
> - rx_ring_lock(ssk, flags);
>
> sdp_rx_ring_destroy(ssk);
> sdp_tx_ring_destroy(ssk);
> @@ -209,9 +207,6 @@ static void sdp_destroy_qp(struct sdp_sock *ssk)
> ib_dealloc_pd(pd);
>
> sdp_remove_large_sock(ssk);
> -
> - rx_ring_unlock(ssk, flags);
> -
> }
>
> static void sdp_reset_keepalive_timer(struct sock *sk, unsigned long len)
> @@ -453,10 +448,6 @@ void sdp_reset_sk(struct sock *sk, int rc)
> sdp_set_error(sk, rc);
> }
>
> - sdp_destroy_qp(ssk);
> -
> - memset((void *)&ssk->id, 0, sizeof(*ssk) - offsetof(typeof(*ssk),
> id));
> -
> sk->sk_state_change(sk);
>
> /* Don't destroy socket before destroy work does its job */
> @@ -976,6 +967,10 @@ void sdp_destroy_work(struct work_struct *work)
> struct sock *sk = &ssk->isk.sk;
> sdp_dbg(sk, "%s: refcnt %d\n", __func__,
> atomic_read(&sk->sk_refcnt));
>
> + sdp_destroy_qp(ssk);
> +
> + memset((void *)&ssk->id, 0, sizeof(*ssk) - offsetof(typeof(*ssk),
> id));
> +
> sdp_cancel_dreq_wait_timeout(ssk);
>
> if (sk->sk_state == TCP_TIME_WAIT)
> @@ -2559,6 +2554,10 @@ static void __exit sdp_exit(void)
> sdp_proc_unregister();
>
> ib_unregister_client(&sdp_client);
> +
> + percpu_counter_destroy(sockets_allocated);
> + percpu_counter_destroy(orphan_count);
> +
> kfree(orphan_count);
> kfree(sockets_allocated);
> }
> diff --git a/drivers/infiniband/ulp/sdp/sdp_rx.c
> b/drivers/infiniband/ulp/sdp/sdp_rx.c
> index 749a83a..38417c4 100644
> --- a/drivers/infiniband/ulp/sdp/sdp_rx.c
> +++ b/drivers/infiniband/ulp/sdp/sdp_rx.c
> @@ -447,12 +447,12 @@ static int sdp_process_rx_ctl_skb(struct sdp_sock
> *ssk, struct sk_buff *skb)
> break;
> case SDP_MID_CHRCVBUF:
> sdp_dbg_data(sk, "Handling RX CHRCVBUF\n");
> - sdp_handle_resize_request(ssk, (struct sdp_chrecvbuf *)h);
> + sdp_handle_resize_request(ssk, (struct sdp_chrecvbuf
> *)(h+1));
> __kfree_skb(skb);
> break;
> case SDP_MID_CHRCVBUF_ACK:
> sdp_dbg_data(sk, "Handling RX CHRCVBUF_ACK\n");
> - sdp_handle_resize_ack(ssk, (struct sdp_chrecvbuf *)h);
> + sdp_handle_resize_ack(ssk, (struct sdp_chrecvbuf *)(h+1));
> __kfree_skb(skb);
> break;
> default:
> @@ -787,9 +787,6 @@ int sdp_rx_ring_create(struct sdp_sock *ssk, struct
> ib_device *device)
> {
> struct ib_cq *rx_cq;
> int rc = 0;
> - unsigned long flags;
> -
> - rx_ring_lock(ssk, flags);
>
> atomic_set(&ssk->rx_ring.head, 1);
> atomic_set(&ssk->rx_ring.tail, 1);
> @@ -797,12 +794,11 @@ int sdp_rx_ring_create(struct sdp_sock *ssk, struct
> ib_device *device)
> ssk->rx_ring.buffer = kmalloc(
> sizeof *ssk->rx_ring.buffer * SDP_RX_SIZE,
> GFP_KERNEL);
> if (!ssk->rx_ring.buffer) {
> - rc = -ENOMEM;
> sdp_warn(&ssk->isk.sk,
> "Unable to allocate RX Ring size %zd.\n",
> sizeof(*ssk->rx_ring.buffer) * SDP_RX_SIZE);
>
> - goto out;
> + return -ENOMEM;
> }
>
> /* TODO: use vector=IB_CQ_VECTOR_LEAST_ATTACHED when implemented
> @@ -822,13 +818,11 @@ int sdp_rx_ring_create(struct sdp_sock *ssk, struct
> ib_device *device)
>
> sdp_arm_rx_cq(&ssk->isk.sk);
>
> - goto out;
> + return 0;
>
> err_cq:
> kfree(ssk->rx_ring.buffer);
> ssk->rx_ring.buffer = NULL;
> -out:
> - rx_ring_unlock(ssk, flags);
> return rc;
> }
>
--
Amir Vadai
Software Eng.
Mellanox Technologies
mailto: amirv at mellanox.co.il
Tel +972-3-6259539
More information about the general
mailing list