Re: [PATCH 2/2] xfs: don't wait on future iclogs when pushing the CIL

[Brian Foster <bfoster@redhat.com>]

On Tue, Jun 15, 2021 at 04:46:58PM +1000, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
> 
> The iclogbuf ring attached to the struct xlog is circular, hence the
> first and last iclogs in the ring can only be determined by
> comparing them against the log->l_iclog pointer.
> 
> In xfs_cil_push_work(), we want to wait on previous iclogs that were
> issued so that we can flush them to stable storage with the commit
> record write, and it simply waits on the previous iclog in the ring.
> This, however, leads to CIL push hangs in generic/019 like so:
> 
> task:kworker/u33:0   state:D stack:12680 pid:    7 ppid:     2 flags:0x00004000
> Workqueue: xfs-cil/pmem1 xlog_cil_push_work
> Call Trace:
>  __schedule+0x30b/0x9f0
>  schedule+0x68/0xe0
>  xlog_wait_on_iclog+0x121/0x190
>  ? wake_up_q+0xa0/0xa0
>  xlog_cil_push_work+0x994/0xa10
>  ? _raw_spin_lock+0x15/0x20
>  ? xfs_swap_extents+0x920/0x920
>  process_one_work+0x1ab/0x390
>  worker_thread+0x56/0x3d0
>  ? rescuer_thread+0x3c0/0x3c0
>  kthread+0x14d/0x170
>  ? __kthread_bind_mask+0x70/0x70
>  ret_from_fork+0x1f/0x30
> 
> With other threads blocking in either xlog_state_get_iclog_space()
> waiting for iclog space or xlog_grant_head_wait() waiting for log
> reservation space.
> 
> The problem here is that the previous iclog on the ring might
> actually be a future iclog. That is, if log->l_iclog points at
> commit_iclog, commit_iclog is the first (oldest) iclog in the ring
> and there are no previous iclogs pending as they have all completed
> their IO and been activated again. IOWs, commit_iclog->ic_prev
> points to an iclog that will be written in the future, not one that
> has been written in the past.
> 
> Hence, in this case, waiting on the ->ic_prev iclog is incorrect
> behaviour, and depending on the state of the future iclog, we can
> end up with a circular ABA wait cycle and we hang.
> 
> Fix this by only waiting on the previous iclog when the commit_iclog
> is not the oldest iclog in the ring.
> 
> Fixes: 5fd9256ce156 ("xfs: separate CIL commit record IO")
> Reported-by: Brian Foster <bfoster@redhat.com>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
>  fs/xfs/xfs_log_cil.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/xfs/xfs_log_cil.c b/fs/xfs/xfs_log_cil.c
> index 705619e9dab4..398f00cf9cbf 100644
> --- a/fs/xfs/xfs_log_cil.c
> +++ b/fs/xfs/xfs_log_cil.c
> @@ -1075,15 +1075,21 @@ xlog_cil_push_work(
>  	ticket = ctx->ticket;
>  
>  	/*
> -	 * If the checkpoint spans multiple iclogs, wait for all previous
> -	 * iclogs to complete before we submit the commit_iclog. In this case,
> -	 * the commit_iclog write needs to issue a pre-flush so that the
> -	 * ordering is correctly preserved down to stable storage.
> +	 * If the checkpoint spans multiple iclogs, wait for all previous iclogs
> +	 * to complete before we submit the commit_iclog. If the commit iclog is
> +	 * at the head of the iclog ring, then all other iclogs have completed
> +	 * and are waiting on this one and hence we don't need to wait.
> +	 *
> +	 * Regardless of whether we need to wait or not, the the commit_iclog
> +	 * write needs to issue a pre-flush so that the ordering for this
> +	 * checkpoint is correctly preserved down to stable storage.
>  	 */
>  	spin_lock(&log->l_icloglock);
>  	if (ctx->start_lsn != commit_lsn) {
> -		xlog_wait_on_iclog(commit_iclog->ic_prev);
> -		spin_lock(&log->l_icloglock);
> +		if (commit_iclog != log->l_iclog) {
> +			xlog_wait_on_iclog(commit_iclog->ic_prev);
> +			spin_lock(&log->l_icloglock);
> +		}

I'm confused at what this is attempting to accomplish. If we have a
single CIL force and that happens to span several iclogs, isn't the goal
to wait on the previous iclog to ensure the previously written content
of the current checkpoint is on-disk? If so, it looks to me that
commit_iclog will always match log->l_iclog in that scenario because in
the commit record path we don't (potentially) switch it out until toward
the end of this function. Hmm.. did you mean to check ->ic_prev against
->l_iclog perhaps, or am I just missing something else?

Brian

>  		commit_iclog->ic_flags |= XLOG_ICL_NEED_FLUSH;
>  	}
>  
> -- 
> 2.31.1
>