LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Milan Svoboda <msvoboda@ra.rockwell.com>
To: linux-kernel@vger.kernel.org
Subject: posix-timers overrun broken?
Date: Wed, 7 Feb 2007 14:36:12 +0100	[thread overview]
Message-ID: <OFA77229BB.3038F286-ONC125727B.0048DC3E-C125727B.004AA20A@ra.rockwell.com> (raw)

Hi,

I'm creating a driver for timer on ixp4xx (there are two hardware timers). 
It shall use posix api
so I looked at the only driver that implements it in kernel: 
drivers/char/mmtimer.c. My driver uses almost the
same logic that handles overruns.

However, I found that receiving of number of overruns in waitsiginfo is 
somewhat broken.

Linux-2.6.20:

Look at posix_timer_event - this is called when interrupt wants to signal 
the thread that
event happend. The very first operation there is
        memset(&timr->sigq->info, 0, sizeof(siginfo_t));
this clears all info members - si.overrun included.
Then send_sigqueue or send_group_sigqueue is called. They both check if 
it's
SI_TIMER and if so, they increase si.overrun if the signal is already 
queued.

But if the next interrupt arrives before function collect_signal is called 
to actually deliver the
siginfo_t to userspace, the si.overrun is cleared in posix_timer_event and 
we have just forgotten
one overrun...

Am I wrong?

Patch follows, but please note that it is more to describe the problem 
than solve the problem in proper
way...

Please, CC me, I'm not subscribed on the list.

Signed-off-by: Milan Svoboda <msvoboda@ra.rockwell.com>

Index: linux/kernel/posix-timers.c
===================================================================
--- linux.orig/kernel/posix-timers.c
+++ linux/kernel/posix-timers.c
@@ -298,7 +298,10 @@ void do_schedule_next_timer(struct sigin
 
 int posix_timer_event(struct k_itimer *timr,int si_private)
 {
+       int overrun = timr->sigq->info.si_overrun;
        memset(&timr->sigq->info, 0, sizeof(siginfo_t));
+       timr->sigq->info.si_overrun = overrun;
+
        timr->sigq->info.si_sys_private = si_private;
        /* Send signal to the process that owns this timer.*/
 
@@ -407,6 +410,7 @@ static struct k_itimer * alloc_posix_tim
                kmem_cache_free(posix_timers_cache, tmr);
                tmr = NULL;
        }
+       tmr->sigq->info.si_overrun = 0;
        return tmr;
 }
 
Index: linux/kernel/signal.c
===================================================================
--- linux.orig/kernel/signal.c
+++ linux/kernel/signal.c
@@ -402,6 +402,7 @@ static int collect_signal(int sig, struc
        if (first) {
                list_del_init(&first->list);
                copy_siginfo(info, &first->info);
+               first->info.si_overrun = 0;
                __sigqueue_free(first);
                if (!still_pending)
                        sigdelset(&list->signal, sig);


             reply	other threads:[~2007-02-07 13:35 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-02-07 13:36 Milan Svoboda [this message]
2007-02-07 16:14 ` Thomas Gleixner
2007-02-08  7:55 Milan Svoboda

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=OFA77229BB.3038F286-ONC125727B.0048DC3E-C125727B.004AA20A@ra.rockwell.com \
    --to=msvoboda@ra.rockwell.com \
    --cc=linux-kernel@vger.kernel.org \
    --subject='Re: posix-timers overrun broken?' \
    /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

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).