Messages in this thread | | | From | OGAWA Hirofumi <> | Subject | Re: [Regression] kdesu broken | Date | Mon, 27 Jul 2009 21:07:28 +0900 |
| |
Alan Cox <alan@lxorguk.ukuu.org.uk> writes:
>> I see. It sounds like good thing. The attached patch or something? >> Anyway, I'm not familiar with the tty stuff obviously, so, I'm not sure >> whether this patch is right or not. > > It turns out to be a little bit trickier than I thought because of open > v close v flush_to_ldisc races (some of the open/close ones being long > standing). > > We now use tty->buf.lock to keep EOF/EOFPENDING/OTHER_CLOSED all in order > together. That has no real cost as we already hold the buf.lock in the hot > path which is flush_to_ldisc. > > Anyway this is my current thoughts (not yet given a testing)
I see. Looks like clean to me.
> + spin_lock_irqsave(&pair->buf.lock, flags); > + pair->packet = 0;
I worried "pair->packet = 0" when I'm thinking this. I guess it would be changed more early than before. Is it ok?
Thanks. -- OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
| |