On Wed 08-05-24 09:42:30, Kemeng Shi wrote:
>
>
> on 5/7/2024 7:51 PM, Zhang Yi wrote:
> > On 2024/5/6 22:17, Kemeng Shi wrote:
> >> The done_copy_out info from jbd2_journal_write_metadata_buffer is not
> >> used. Simply remove it.
> >>
> >> Signed-off-by: Kemeng Shi <shikemeng@xxxxxxxxxxxxxxx>
> >> ---
> >> fs/jbd2/journal.c | 3 +--
> >> 1 file changed, 1 insertion(+), 2 deletions(-)
> >>
> >> diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
> >> index 207b24e12ce9..068031f35aea 100644
> >> --- a/fs/jbd2/journal.c
> >> +++ b/fs/jbd2/journal.c
> >> @@ -320,7 +320,6 @@ static void journal_kill_thread(journal_t *journal)
> >> *
> >> * On success:
> >> * Bit 0 set == escape performed on the data
> >> - * Bit 1 set == buffer copy-out performed (kfree the data after IO)
> >> */
> >>
> >> int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
> >> @@ -455,7 +454,7 @@ int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
> >> set_buffer_shadow(bh_in);
> >> spin_unlock(&jh_in->b_state_lock);
> >>
> >> - return do_escape | (done_copy_out << 1);
> >> + return do_escape;
> >> }
> >>
> >
> > I checked the history, it seems that this bit has not been used since
> > the very beginning when the jbd code was merged in, I suppose we could
> > drop it. Since there is only one flag that is still in use, why not just
> > drop the flag and pass out do_escape through an output parameter, or
> > directly pass tag_flag, after that we could also drop the weird
> > "if (flags & 1)" check in jbd2_journal_commit_transaction()?
> Thanks for looking into this. I agree that the "flags & 1" is wired, I
> wonder if we could further remove "flags & 1" with following change:
>
> diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
> index 5e122586e06e..67077308b56b 100644
> --- a/fs/jbd2/commit.c
> +++ b/fs/jbd2/commit.c
> @@ -353,7 +353,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
> struct buffer_head *descriptor;
> struct buffer_head **wbuf = journal->j_wbuf;
> int bufs;
> - int flags;
> + int escape;
> int err;
> unsigned long long blocknr;
> ktime_t start_time;
> @@ -661,10 +661,10 @@ void jbd2_journal_commit_transaction(journal_t *journal)
> */
> set_bit(BH_JWrite, &jh2bh(jh)->b_state);
> JBUFFER_TRACE(jh, "ph3: write metadata");
> - flags = jbd2_journal_write_metadata_buffer(commit_transaction,
> + escape = jbd2_journal_write_metadata_buffer(commit_transaction,
> jh, &wbuf[bufs], blocknr);
> - if (flags < 0) {
> - jbd2_journal_abort(journal, flags);
> + if (escape < 0) {
> + jbd2_journal_abort(journal, escape);
> continue;
> }
> jbd2_file_log_bh(&io_bufs, wbuf[bufs]);
> @@ -673,7 +673,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
> buffer */
>
> tag_flag = 0;
> - if (flags & 1)
> + if (escape)
> tag_flag |= JBD2_FLAG_ESCAPE;
> if (!first_tag)
> tag_flag |= JBD2_FLAG_SAME_UUID;
> diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
> index 08c59197c5bd..c8a1eca5caab 100644
> --- a/fs/jbd2/journal.c
> +++ b/fs/jbd2/journal.c
> @@ -309,10 +309,8 @@ static void journal_kill_thread(journal_t *journal)
> *
> * Return value:
> * <0: Error
> - * >=0: Finished OK
> - *
> - * On success:
> - * Bit 0 set == escape performed on the data
> + * =0: Finished OK without escape
> + * =1: Finished OK with escape
> */
>
> Look forward to your reply.
Yep, looks like a fine solution to me.
Honza
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
