Date: Sun, 27 Apr 2003 22:35:06 -0700 (PDT) From: Marcel Moolenaar <marcel@FreeBSD.org> To: Perforce Change Reviews <perforce@freebsd.org> Subject: PERFORCE change 29931 for review Message-ID: <200304280535.h3S5Z6uc081449@repoman.freebsd.org>
next in thread | raw e-mail | index | archive | help
http://perforce.freebsd.org/chv.cgi?CH=29931 Change 29931 by marcel@marcel_nfs on 2003/04/27 22:34:48 We have an annoying race when using EPC syscalls which results in a general exception in cpu_switch(). We obviously get an interrupt when setting up the trapframe, but why and how this result in cloberring is not know yet. So, apply the equivalent of a 9-inch nail and disable interrupts. I need to understand what is happening exactly so that I can create a real fix (or acknowledge that this is in fact the fix). For now it simply makes EPC syscalls reliable enough to make the switch. Affected files ... .. //depot/projects/ia64_epc/sys/ia64/ia64/syscall.s#13 edit Differences ... ==== //depot/projects/ia64_epc/sys/ia64/ia64/syscall.s#13 (text+ko) ==== @@ -224,6 +224,8 @@ .section .text.syscall, "ax" ENTRY(epc_syscall, 8) + rsm psr.i + ;; { .mmi mov r16=ar.rsc mov ar.rsc=0 @@ -359,7 +361,7 @@ ;; } { .mlx - ssm psr.dfh|psr.ac + ssm psr.dfh|psr.ac|psr.i movl gp=__gp ;; }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200304280535.h3S5Z6uc081449>