linux-xfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Brian Foster <bfoster@redhat.com>
To: Dave Chinner <david@fromorbit.com>
Cc: linux-xfs@vger.kernel.org
Subject: Re: [PATCH 01/11] xfs: don't try to write a start record into every iclog
Date: Thu, 5 Mar 2020 13:05:31 -0500	[thread overview]
Message-ID: <20200305180531.GA28340@bfoster> (raw)
In-Reply-To: <20200304075401.21558-2-david@fromorbit.com>

On Wed, Mar 04, 2020 at 06:53:51PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
> 
> The xlog_write() function iterates over iclogs until it completes
> writing all the log vectors passed in. The ticket tracks whether
> a start record has been written or not, so only the first iclog gets
> a start record. We only ever pass single use tickets to
> xlog_write() we only ever need to write a start record once per

	      ^ "so we only ever ..." ?

> xlog_write() call.
> 
> Hence we don't need to store whether we should write a start record
> in the ticket as the callers provide all the information we need to
> determine if a start record should be written. For the moment, we
> have to ensure that we clear the XLOG_TIC_INITED appropriately so
> the code in xfs_log_done() still works correctly for committing
> transactions.
> 
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
>  fs/xfs/xfs_log.c | 68 +++++++++++++++++++++++++-----------------------
>  1 file changed, 35 insertions(+), 33 deletions(-)
> 
> diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
> index f6006d94a581..5b0568a86c07 100644
> --- a/fs/xfs/xfs_log.c
> +++ b/fs/xfs/xfs_log.c
...
> @@ -2404,25 +2391,29 @@ xlog_write(
>  	int			record_cnt = 0;
>  	int			data_cnt = 0;
>  	int			error = 0;
> +	int			start_rec_size = sizeof(struct xlog_op_header);
>  
>  	*start_lsn = 0;
>  
> -	len = xlog_write_calc_vec_length(ticket, log_vector);
>  
>  	/*
>  	 * Region headers and bytes are already accounted for.
>  	 * We only need to take into account start records and
>  	 * split regions in this function.
>  	 */
> -	if (ticket->t_flags & XLOG_TIC_INITED)
> +	if (ticket->t_flags & XLOG_TIC_INITED) {
>  		ticket->t_curr_res -= sizeof(xlog_op_header_t);
> +		ticket->t_flags &= ~XLOG_TIC_INITED;
> +	}
>  
>  	/*
>  	 * Commit record headers need to be accounted for. These
>  	 * come in as separate writes so are easy to detect.
>  	 */
> -	if (flags & (XLOG_COMMIT_TRANS | XLOG_UNMOUNT_TRANS))
> +	if (flags & (XLOG_COMMIT_TRANS | XLOG_UNMOUNT_TRANS)) {
>  		ticket->t_curr_res -= sizeof(xlog_op_header_t);
> +		start_rec_size = 0;
> +	}

This is more of a flaw in the existing code than this patch, but please
update the comment to document why we reset start_rec_size here since
it's not obvious from the context of the function. Otherwise looks Ok to
me modulo Christoph's comments (in particular the data_cnt bit). My one
nit is that I think start_rec might be a better name than start_rec_size
given how it's used, but I don't care that much.

Brian

>  
>  	if (ticket->t_curr_res < 0) {
>  		xfs_alert_tag(log->l_mp, XFS_PTAG_LOGRES,
> @@ -2431,6 +2422,8 @@ xlog_write(
>  		xfs_force_shutdown(log->l_mp, SHUTDOWN_LOG_IO_ERROR);
>  	}
>  
> +	len = xlog_write_calc_vec_length(ticket, log_vector, start_rec_size);
> +
>  	index = 0;
>  	lv = log_vector;
>  	vecp = lv->lv_iovecp;
> @@ -2446,9 +2439,20 @@ xlog_write(
>  		ASSERT(log_offset <= iclog->ic_size - 1);
>  		ptr = iclog->ic_datap + log_offset;
>  
> -		/* start_lsn is the first lsn written to. That's all we need. */
> +		/*
> +		 * Before we start formatting log vectors, we need to write a
> +		 * start record and record the lsn of the iclog that we write
> +		 * the start record into. Only do this for the first iclog we
> +		 * write to.
> +		 */
>  		if (!*start_lsn)
>  			*start_lsn = be64_to_cpu(iclog->ic_header.h_lsn);
> +		if (start_rec_size) {
> +			xlog_write_start_rec(ptr, ticket);
> +			xlog_write_adv_cnt(&ptr, &len, &log_offset,
> +					start_rec_size);
> +			record_cnt++;
> +		}
>  
>  		/*
>  		 * This loop writes out as many regions as can fit in the amount
> @@ -2457,7 +2461,6 @@ xlog_write(
>  		while (lv && (!lv->lv_niovecs || index < lv->lv_niovecs)) {
>  			struct xfs_log_iovec	*reg;
>  			struct xlog_op_header	*ophdr;
> -			int			start_rec_copy;
>  			int			copy_len;
>  			int			copy_off;
>  			bool			ordered = false;
> @@ -2473,13 +2476,6 @@ xlog_write(
>  			ASSERT(reg->i_len % sizeof(int32_t) == 0);
>  			ASSERT((unsigned long)ptr % sizeof(int32_t) == 0);
>  
> -			start_rec_copy = xlog_write_start_rec(ptr, ticket);
> -			if (start_rec_copy) {
> -				record_cnt++;
> -				xlog_write_adv_cnt(&ptr, &len, &log_offset,
> -						   start_rec_copy);
> -			}
> -
>  			ophdr = xlog_write_setup_ophdr(log, ptr, ticket, flags);
>  			if (!ophdr)
>  				return -EIO;
> @@ -2509,10 +2505,15 @@ xlog_write(
>  				xlog_write_adv_cnt(&ptr, &len, &log_offset,
>  						   copy_len);
>  			}
> -			copy_len += start_rec_copy + sizeof(xlog_op_header_t);
> +			copy_len += sizeof(xlog_op_header_t);
>  			record_cnt++;
>  			data_cnt += contwr ? copy_len : 0;
>  
> +			if (start_rec_size) {
> +				copy_len += start_rec_size;
> +				start_rec_size = 0;
> +			}
> +
>  			error = xlog_write_copy_finish(log, iclog, flags,
>  						       &record_cnt, &data_cnt,
>  						       &partial_copy,
> @@ -2550,6 +2551,7 @@ xlog_write(
>  				break;
>  			}
>  		}
> +		start_rec_size = 0;
>  	}
>  
>  	ASSERT(len == 0);
> -- 
> 2.24.0.rc0
> 


  parent reply	other threads:[~2020-03-05 18:05 UTC|newest]

Thread overview: 44+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-04  7:53 [PATCH 00/11] xfs: clean up log tickets and record writes Dave Chinner
2020-03-04  7:53 ` [PATCH 01/11] xfs: don't try to write a start record into every iclog Dave Chinner
2020-03-04 15:44   ` Christoph Hellwig
2020-03-04 21:26     ` Dave Chinner
2020-03-05 15:19       ` Christoph Hellwig
2020-03-05 18:05   ` Brian Foster [this message]
2020-03-04  7:53 ` [PATCH 02/11] xfs: re-order initial space accounting checks in xlog_write Dave Chinner
2020-03-05 18:05   ` Brian Foster
2020-03-04  7:53 ` [PATCH 03/11] xfs: refactor and split xfs_log_done() Dave Chinner
2020-03-04 15:49   ` Christoph Hellwig
2020-03-04 21:28     ` Dave Chinner
2020-03-05 15:20       ` Christoph Hellwig
2020-03-05 18:06   ` Brian Foster
2020-03-24 12:37     ` Christoph Hellwig
2020-03-04  7:53 ` [PATCH 04/11] xfs: merge xlog_commit_record with xlog_write_done() Dave Chinner
2020-03-04 15:50   ` Christoph Hellwig
2020-03-05 18:06   ` Brian Foster
2020-03-04  7:53 ` [PATCH 05/11] xfs: factor out unmount record writing Dave Chinner
2020-03-04 15:51   ` Christoph Hellwig
2020-03-05 18:07   ` Brian Foster
2020-03-04  7:53 ` [PATCH 06/11] xfs: move xlog_state_ioerror() Dave Chinner
2020-03-04 15:51   ` Christoph Hellwig
2020-03-04 21:41     ` Dave Chinner
2020-03-05 15:21       ` Christoph Hellwig
2020-03-05 18:07   ` Brian Foster
2020-03-04  7:53 ` [PATCH 07/11] xfs: clean up xlog_state_ioerror() Dave Chinner
2020-03-05 18:07   ` Brian Foster
2020-03-04  7:53 ` [PATCH 08/11] xfs: rename the log unmount writing functions Dave Chinner
2020-03-04 15:52   ` Christoph Hellwig
2020-03-05 18:07   ` Brian Foster
2020-03-04  7:53 ` [PATCH 09/11] xfs: merge unmount record write iclog cleanup Dave Chinner
2020-03-04 15:53   ` Christoph Hellwig
2020-03-04 21:38     ` Dave Chinner
2020-03-05 15:27       ` Christoph Hellwig
2020-03-05 18:08   ` Brian Foster
2020-03-04  7:54 ` [PATCH 10/11] xfs: remove some stale comments from the log code Dave Chinner
2020-03-04 15:53   ` Christoph Hellwig
2020-03-05 18:08   ` Brian Foster
2020-03-04  7:54 ` [PATCH 11/11] xfs: kill XLOG_TIC_INITED Dave Chinner
2020-03-04 15:54   ` Christoph Hellwig
2020-03-05 18:08   ` Brian Foster
2020-03-05 16:05 ` [PATCH 00/11] xfs: clean up log tickets and record writes Christoph Hellwig
2020-03-05 21:42   ` Dave Chinner
2020-03-06  1:03     ` Christoph Hellwig

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200305180531.GA28340@bfoster \
    --to=bfoster@redhat.com \
    --cc=david@fromorbit.com \
    --cc=linux-xfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).