Re: O_SYNC patch -> slow fb memory access

2003-09-21 Thread Thomas Winischhofer
Branden Robinson wrote: On Sun, Sep 21, 2003 at 10:47:57AM +0200, Egbert Eich wrote: No, I removed it again. O_SYNC causes the memory to be mapped uncached. I had a system which crashed horribly when this flag was not set for MMIO. Now I believe the system itself was faulty. I'm still not sure

Re: O_SYNC patch -> slow fb memory access

2003-09-21 Thread Thomas Winischhofer
Branden Robinson wrote: On Sun, Sep 21, 2003 at 10:47:57AM +0200, Egbert Eich wrote: No, I removed it again. O_SYNC causes the memory to be mapped uncached. I had a system which crashed horribly when this flag was not set for MMIO. Now I believe the system itself was faulty. I'm still not sure if

Re: O_SYNC patch -> slow fb memory access

2003-09-21 Thread Branden Robinson
On Sun, Sep 21, 2003 at 10:47:57AM +0200, Egbert Eich wrote: > No, I removed it again. O_SYNC causes the memory to be mapped > uncached. I had a system which crashed horribly when this flag > was not set for MMIO. Now I believe the system itself was faulty. > I'm still not sure if caching should be

Re: O_SYNC patch -> slow fb memory access

2003-09-21 Thread Branden Robinson
On Sun, Sep 21, 2003 at 10:47:57AM +0200, Egbert Eich wrote: > No, I removed it again. O_SYNC causes the memory to be mapped > uncached. I had a system which crashed horribly when this flag > was not set for MMIO. Now I believe the system itself was faulty. > I'm still not sure if caching should be

Re: O_SYNC patch -> slow fb memory access

2003-09-21 Thread Egbert Eich
Thomas Winischhofer writes: > > Egbert, Debian folks: > > After the release of XFree86 4.3.0, (AFAIK) Egbert Eich commited a patch > which set the O_SYNC flag when opening /dev/mem for framebuffer memory > access. > > This caused memory access to the framebuffer area to become terribly

Re: O_SYNC patch -> slow fb memory access

2003-09-21 Thread Egbert Eich
Thomas Winischhofer writes: > > Egbert, Debian folks: > > After the release of XFree86 4.3.0, (AFAIK) Egbert Eich commited a patch > which set the O_SYNC flag when opening /dev/mem for framebuffer memory > access. > > This caused memory access to the framebuffer area to become terribly

O_SYNC patch -> slow fb memory access

2003-09-20 Thread Thomas Winischhofer
Egbert, Debian folks: After the release of XFree86 4.3.0, (AFAIK) Egbert Eich commited a patch which set the O_SYNC flag when opening /dev/mem for framebuffer memory access. This caused memory access to the framebuffer area to become terribly slow on some machines. Among these types of mach

O_SYNC patch -> slow fb memory access

2003-09-20 Thread Thomas Winischhofer
Egbert, Debian folks: After the release of XFree86 4.3.0, (AFAIK) Egbert Eich commited a patch which set the O_SYNC flag when opening /dev/mem for framebuffer memory access. This caused memory access to the framebuffer area to become terribly slow on some machines. Among these types of machine