talexander
/
freebsd_amp_hwpstate
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Sigh. Signal handlers are executed on leaving the system call, not 

at moment of delivery.  Restoring the signal mask after the tsleep()
is next to useless since the signal is still queued.. This was interacting
with usleep(3) on receipt of a SIGALRM causing it to near busy loop.

Now, we set the new signal mask "permanently" for signanosleep().

Problem noted by:  bde
This commit is contained in:
Peter Wemm 1997-10-15 13:58:52 +00:00
parent 138ec1f71a
commit 987163643c
1 changed files with 2 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
sys/kern/kern_time.c
View File

@ -31,7 +31,7 @@
* SUCH DAMAGE.
*
* @(#)kern_time.c 8.1 (Berkeley) 6/10/93
* $Id: kern_time.c,v 1.34 1997/09/02 20:05:49 bde Exp $
* $Id: kern_time.c,v 1.35 1997/09/21 22:00:16 gibbs Exp $
*/
#include <sys/param.h>
@ -342,17 +342,11 @@ signanosleep(p, uap, retval)
if (error)
return (error);
/* See kern_sig.c:sigsuspend() for explanation. */
ps->ps_oldmask = p->p_sigmask;
ps->ps_flags |= SAS_OLDMASK;
/* change mask for sleep */
p->p_sigmask = mask &~ sigcantmask;
error = nanosleep1(p, &rqt, &rmt);
/* See kern_sig.c:sigsuspend() again. */
p->p_sigmask = ps->ps_oldmask; /* in case timeout rather than sig */
ps->ps_flags &= ~SAS_OLDMASK;
if (SCARG(uap, rmtp)) {
error2 = copyout(&rmt, SCARG(uap, rmtp), sizeof(rmt));
if (error2) /* XXX shouldn't happen, did useracc() above */