Date: Tue, 02 Jun 2020 07:39:48 -0500 From: Scott Bennett <bennett@sdf.org> To: ml@netfence.it Cc: freebsd-x11@freebsd.org Subject: Re: xf86-video-ati-legacy port status Message-ID: <202006021239.052CdmTm000452@sdf.org> In-Reply-To: <14a0ce02-6c8c-e9e6-2e9f-61069e177459@netfence.it> References: <20200527172918.GA874@doom.homeunix.org> <ab0ff124-14ca-83a8-8643-65a0a819cf42@freebsd.org> <6675586c-dae2-39c6-2aba-bd66511d2d72@netfence.it> <20200601211504.72eab297bca88aacb165b21c@bidouilliste.com> <a02dd18b-a14e-700b-d7b6-b5ba3954bcd6@netfence.it> <202006020912.0529C4Hg025076@sdf.org> <14a0ce02-6c8c-e9e6-2e9f-61069e177459@netfence.it>
index | next in thread | previous in thread | raw e-mail
Andrea Venturoli <ml@netfence.it> wrote: > On 2020-06-02 11:12, Scott Bennett wrote: > > >> https://lists.freebsd.org/pipermail/freebsd-x11/2020-May/025989.html > > > Perhaps you posted your panic messages, and I missed them, > > Link is above. > Oh, I see. Okay, that is something quite different. > > > but I've > > been wondering whether you actually get a panic screen with all sorts of > > kernel panic messages, > > Yes: the system drops into DDB. > > > or do you instead get a hard BIOS reset > > No. > > > in an attempt by the DRM driver to reboot your system? > > Here I'm not so sure... I guess not, since what I get is "panic: page > fault". > Right. > > > The reason I ask is that > > the latter is what happens when running xorg on a Radeon HD 5770, and after > > a reboot I can see the final messages that were written to /var/log/messages > > just before the reset. Those messages indicated that the GPU had crashed > > (bug #1). The improper response by the DRM driver is bug #2/design flaw #1, > > depending upon how you view such things, although a design document might > > well resolve which sort of error was made. > > I don't seem to follow you here. > Are those # references to some bug list? Not at all. I was just pointing out that there appear to be two distinct bugs involved that make running xorg on a Radeon HD 5770 very dangerous to production operations. The numbering merely reflects the order in which they occur in crashing my system. Bug #1 is probably in a gpu-firmware-kmod module loaded by drm-fbsd11.2-kmod. Bug #2/design flaw #1 happens second and is the decision to crash the system, rather than reinitialize the GPU when it has hung or crashed. Bug #1 has thus far always occurred within 36 hours of starting xorg, and because bug #2 means the response to bug #1, on my system it always results in lost mprime work and taking down my tor relay. The latter means the relay loses its Stable flag status (determined hourly by the directory authorities), which isn't regained until the relay has remained up and running for several days again. So my choices are run tor or run xorg, dealing with the crashes and the nuisance of cleaning up the wreckage every time. Because I don't dare run xorg, I have no way to file PRs, so the problems never get addressed. It's a nice way for the graphics team to shrug off a pair of bugs without doing anything to fix them. BTW, you needn't insert all those extraneous blank lines in your postings. I deleted a bunch of them in this followup. Scotthome | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202006021239.052CdmTm000452>