[Hardware] насчет swsusp + i855GM [Suspending with i915 and DRI (possible solution?)]

[Konstantin A. Lepikhov]

Hi!

JFYI как возможный workaround (хотя надо смотреть патчи для xorg).

----- Forwarded message from Bernard Blackham <bernard на blackham.com.au> -----

Date: Fri, 7 Jan 2005 22:12:25 +0800
From: Bernard Blackham <bernard@>
To: dri-devel@
Subject: Suspending with i915 and DRI (possible solution?)
Organization: Dagobah Systems
User-Agent: Mutt/1.5.6+20040907i

Greetings,

I have an Intel i855GM chipset in my laptop, running Xorg 6.8.1 with
i810 and i915_dri for DRI, and kernel 2.6.10 with the i915 kernel
module. DRI is all well and happy, and glxgears achieves about
800fps. I use Software Suspend 2 (2.1.5.11) from
softwaresuspend.berlios.de, and upon suspending and resuming, I
found that GL apps would freeze when trying to be launched (as has
been reported before).

So I did some digging and poking, and wrote the attached little C
program that calls I915_RESUME_DMA in the i915 drm driver, and
surprisingly, this lets me run GL applications once more without
restarting X.

If I simply run it after resuming and returning back to X, I can run
GL applications again. Even more interestingly though, is if I run
it after resuming but before returning back to X, existing GL
applications don't lock up and continue to run, with some small
mishaps:

 - on the first suspend invocation (from a freshly started X), Xorg
   will consume 99% CPU (though still be otherwise functional) until
   I kill glxgears *and* switch VTs to text and back to X. On
   subsequent suspend invocations, this doesn't happen. If I'm not
   running glxgears or a GL app when suspending, it doesn't happen
   either.

 - glxgears' reported fps will drop from the usual 800fps on a
   freshly started X, to about 200fps on a suspend/resumed one, but
   CPU usage is pretty close to zero indicating that it's certainly
   not software-GL. I can still play tuxracer after resume with no
   noticable differences.

 - running tuxracer whilst suspending and resuming results in a
   garbled screen (fair enough though :). You can make out enough of
   the screen to exit tuxracer and restart it again... I guess I'll
   have to stop tuxracer whilst suspending for now ;)

So this is really just a quick hack but is it worth considering
moving into something like the X dri driver? (eg, on vt switch?)

And is it just working for me out of pure luck? :)

I can imagine that to restore full DRI would take much more work
(textures and the like).

TIA,

Bernard.

-- 
 Bernard Blackham <bernard at blackham dot com dot au>

#include <sys/ioctl.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

#define DRM_IOCTL_I915_INIT 0x40

typedef struct _drm_i915_init {
    enum {   
	I915_INIT_DMA = 0x01,
	I915_CLEANUP_DMA = 0x02,
	I915_RESUME_DMA = 0x03
    } func;
    unsigned int mmio_offset;
    int sarea_priv_offset;
    unsigned int ring_start;
    unsigned int ring_end;
    unsigned int ring_size;
    unsigned int front_offset;
    unsigned int back_offset;
    unsigned int depth_offset;
    unsigned int w;
    unsigned int h;
    unsigned int pitch;
    unsigned int pitch_bits;
    unsigned int back_pitch;
    unsigned int depth_pitch;
    unsigned int cpp;
    unsigned int chipset;
} drm_i915_init_t;

int main() {
    drm_i915_init_t init;
    init.func = I915_RESUME_DMA;
    int fd = open("/dev/dri/card0", O_RDWR);
    if (fd == -1) {
	perror("open");
	return 1;
    }
    if (ioctl(fd, DRM_IOCTL_I915_INIT, &init) == -1) {
	perror("ioctl");
	return 1;
    }
    return 0;
}


----- End forwarded message -----

-- 
WBR, Konstantin	      chat with ==>ICQ: 109916175
     Lepikhov,	      speak  to ==>JID: lakostis на jabber.org
aka L.A. Kostis       write  to ==>mailto:lakostis на pisem.net.nospam

...The information is like the bank... 			  (c) EC8OR