queue-6.8/sunrpc-increase-size-of-rpc_wait_queue.qlen-from-uns.patch

   1 From 4b4cfeefb2227540e2d5f5bce7aa2a4d89701e84 Mon Sep 17 00:00:00 2001
   2 From: Sasha Levin <sashal@kernel.org>
   3 Date: Tue, 30 Jan 2024 11:38:25 -0800
   4 Subject: SUNRPC: increase size of rpc_wait_queue.qlen from unsigned short to
   5  unsigned int
   6
   7 From: Dai Ngo <dai.ngo@oracle.com>
   8
   9 [ Upstream commit 2c35f43b5a4b9cdfaa6fdd946f5a212615dac8eb ]
  10
  11 When the NFS client is under extreme load the rpc_wait_queue.qlen counter
  12 can be overflowed. Here is an instant of the backlog queue overflow in a
  13 real world environment shown by drgn helper:
  14
  15 rpc_task_stats(rpc_clnt):
  16 -------------------------
  17 rpc_clnt: 0xffff92b65d2bae00
  18 rpc_xprt: 0xffff9275db64f000
  19   Queue:  sending[64887] pending[524] backlog[30441] binding[0]
  20 XMIT task: 0xffff925c6b1d8e98
  21      WRITE: 750654
  22         __dta_call_status_580: 65463
  23         __dta_call_transmit_status_579: 1
  24         call_reserveresult: 685189
  25         nfs_client_init_is_complete: 1
  26     COMMIT: 584
  27         call_reserveresult: 573
  28         __dta_call_status_580: 11
  29     ACCESS: 1
  30         __dta_call_status_580: 1
  31    GETATTR: 10
  32         __dta_call_status_580: 4
  33         call_reserveresult: 6
  34 751249 tasks for server 111.222.333.444
  35 Total tasks: 751249
  36
  37 count_rpc_wait_queues(xprt):
  38 ----------------------------
  39 **** rpc_xprt: 0xffff9275db64f000 num_reqs: 65511
  40 wait_queue: xprt_binding[0] cnt: 0
  41 wait_queue: xprt_binding[1] cnt: 0
  42 wait_queue: xprt_binding[2] cnt: 0
  43 wait_queue: xprt_binding[3] cnt: 0
  44 rpc_wait_queue[xprt_binding].qlen: 0 maxpriority: 0
  45 wait_queue: xprt_sending[0] cnt: 0
  46 wait_queue: xprt_sending[1] cnt: 64887
  47 wait_queue: xprt_sending[2] cnt: 0
  48 wait_queue: xprt_sending[3] cnt: 0
  49 rpc_wait_queue[xprt_sending].qlen: 64887 maxpriority: 3
  50 wait_queue: xprt_pending[0] cnt: 524
  51 wait_queue: xprt_pending[1] cnt: 0
  52 wait_queue: xprt_pending[2] cnt: 0
  53 wait_queue: xprt_pending[3] cnt: 0
  54 rpc_wait_queue[xprt_pending].qlen: 524 maxpriority: 0
  55 wait_queue: xprt_backlog[0] cnt: 0
  56 wait_queue: xprt_backlog[1] cnt: 685801
  57 wait_queue: xprt_backlog[2] cnt: 0
  58 wait_queue: xprt_backlog[3] cnt: 0
  59 rpc_wait_queue[xprt_backlog].qlen: 30441 maxpriority: 3 [task cnt mismatch]
  60
  61 There is no effect on operations when this overflow occurs. However
  62 it causes confusion when trying to diagnose the performance problem.
  63
  64 Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
  65 Reviewed-by: Jeff Layton <jlayton@kernel.org>
  66 Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
  67 Signed-off-by: Sasha Levin <sashal@kernel.org>
  68 ---
  69  include/linux/sunrpc/sched.h | 2 +-
  70  1 file changed, 1 insertion(+), 1 deletion(-)
  71
  72 diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h
  73 index 2d61987b35456..0c77ba488bbae 100644
  74 --- a/include/linux/sunrpc/sched.h
  75 +++ b/include/linux/sunrpc/sched.h
  76 @@ -197,7 +197,7 @@ struct rpc_wait_queue {
  77         unsigned char           maxpriority;            /* maximum priority (0 if queue is not a priority queue) */
  78         unsigned char           priority;               /* current priority */
  79         unsigned char           nr;                     /* # tasks remaining for cookie */
  80 -       unsigned short          qlen;                   /* total # tasks waiting in queue */
  81 +       unsigned int            qlen;                   /* total # tasks waiting in queue */
  82         struct rpc_timer        timer_list;
  83  #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) || IS_ENABLED(CONFIG_TRACEPOINTS)
  84         const char *            name;
  85 --
  86 2.43.0
  87