Date: Wed, 1 May 2002 20:41:43 +0100 From: Doug Rabson <dfr@nlsystems.com> To: Marcel Moolenaar <marcel@xcllnt.net> Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/ia64/ia64 exception.s src/sys/ia64/include frame.h Message-ID: <200205012041.43577.dfr@nlsystems.com> In-Reply-To: <20020501193719.GA699@athlon.pn.xcllnt.net> References: <200204300637.g3U6bQc25956@freefall.freebsd.org> <200205012015.32377.dfr@nlsystems.com> <20020501193719.GA699@athlon.pn.xcllnt.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday 01 May 2002 8:37 pm, Marcel Moolenaar wrote: > On Wed, May 01, 2002 at 08:15:32PM +0100, Doug Rabson wrote: > > > > What I actually meant by that statement was that since the except= ion > > > > entry path didn't disturb the value of ar.lc (or ar.ec for that > > > > matter) and that it can rely on the value being preserved by the > > > > higher level handlers (e.g. trap, interrupt, syscall etc), it can > > > > also rely on the value in > > > > exception_restore being the user's value. > > > > > > That is correct, provided we save them on context switches, right? > > > > Right. > > Ok, I'll add them to savectx() and cpu_switch() then :-) > > > > BTW: The reason I added them was for ptrace(2). > > > > Thats as good a reason as any. Another way would be to use the unwind > > system to find if/where the register had been saved and access it eit= her > > directly or at the save location. > > Wicked! I'll let this run through my head for a while. My first reactio= n > is that it sounds like a very costly approach, time and complexity wise= =2E > For example: if ar.lc has been saved in the process' pcb, then there's > no way the unwind information can tell us that. We need to handle those > cases seperately, AFAICT. Actually, it would probably be possible to write some carefully crafted u= nwind=20 records in exception_save which describe the locations where stuff gets p= ut=20 in the trapframe... Its certainly a costly approach but for rarely-accessed state, it might b= e=20 worth it compared to the alternative (adding *everything* to struct=20 trapframe). --=20 Doug Rabson=09=09=09=09Mail: dfr@nlsystems.com =09=09=09=09=09Phone: +44 20 8348 6160 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200205012041.43577.dfr>