[Intel-gfx] [PATCH igt] igt: Add basic framework for GVT-g testing

[Chris Wilson]

On Tue, Jun 21, 2016 at 01:24:26PM +0000, Wang, Zhi A wrote:
> > +	while ((vtcon = readdir(dir))) {
> > +		int fd, len;
> > +
> > +		if (strncmp(vtcon->d_name, "vtcon", 5))
> > +			continue;
> > +
> > +		sprintf(buf, "%s/%s/name", path, vtcon->d_name);
> > +		fd = open(buf, O_RDONLY);
> > +		if (fd < 0)
> > +			continue;
> > +
> > +		len = read(fd, buf, sizeof(buf) - 1);
> Will i-g-t use SIGALRM in global test framework? If yes, maybe we should check len == -1 && errno == -EINTR or set SA_RESTART in sigaction(SIGALRM).

Interrupts tend to be added on demand, If we are being thorough, then
yes we should do a readN / writeN here. (Not that I think either kernfs
here reports EINTR.)

> > +static void unload_i915(void)
> > +{
> > +	unbind_fbcon();
> > +	/* pkill alsact */
> > +
> 
> Is there any reason that we should care about alsact? Are you concerned about sound card configuration tool will cause i915 unload failure? BTW I think maybe it should be alsactl?

Yes, the sound driver can keep i915 loaded, so we need to kill anything
keeping the snd_hda_intel alive before we can (recursively) remove i915.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre