Date: Fri, 11 Dec 2009 12:37:17 -0600 From: Robert Noland <rnoland@FreeBSD.org> To: Steve Polyack <korvus@comcast.net> Cc: freebsd-x11 <freebsd-x11@freebsd.org>, John Baldwin <jhb@freebsd.org> Subject: Re: PCI Radeon 9250 - DRI/DRM in 8.0-RELEASE Message-ID: <1260556637.2281.19.camel@balrog.2hip.net> In-Reply-To: <1260476369.2281.16.camel@balrog.2hip.net> References: <4B213D8F.6080906@comcast.net> <1260474623.2281.8.camel@balrog.2hip.net> <4B215405.2080502@comcast.net> <1260476369.2281.16.camel@balrog.2hip.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--=-YktE7EncAO1fO72SX6x0 Content-Type: text/plain Content-Transfer-Encoding: 7bit On Thu, 2009-12-10 at 14:19 -0600, Robert Noland wrote: > On Thu, 2009-12-10 at 15:03 -0500, Steve Polyack wrote: > > Robert Noland wrote: > > > On Thu, 2009-12-10 at 13:27 -0500, Steve Polyack wrote: > > > > > >> (II) RADEON(0): [pci] 8192 kB allocated with handle 0xe9b02000 > > >> (II) RADEON(0): [pci] ring handle = 0xe9b02000 > > >> (II) RADEON(0): [pci] Ring mapped at 0x28a7d000 > > >> (II) RADEON(0): [pci] Ring contents 0x00eeeeec > > >> (II) RADEON(0): [pci] ring read ptr handle = 0xe9c03000 > > >> (II) RADEON(0): [pci] Ring read ptr mapped at 0x286ff000 > > >> (II) RADEON(0): [pci] Ring read ptr contents 0x00ffffff > > >> (II) RADEON(0): [pci] vertex/indirect buffers handle = 0xe9c04000 > > >> (II) RADEON(0): [pci] Vertex/indirect buffers mapped at 0x30c00000 > > >> (II) RADEON(0): [pci] Vertex/indirect buffers contents 0x00ffffff > > >> (II) RADEON(0): [pci] GART texture map handle = 0xe9e04000 > > >> (II) RADEON(0): [pci] GART Texture map mapped at 0x30e00000 > > >> (II) RADEON(0): [drm] register handle = 0xfe5d0000 > > >> > > > > > > > > > The corruption is due to the ring not being zeroed as it should be. > > > This seems to only occur on i386 builds and even then, only sometimes. > > > I'm trying to look into this issue again and I've previously eliminated > > > everything except libdrm and the kernel from this issue by testing with > > > a small sample program that directly interacts with drm without X > > > running. All of the pointers seem to be correct and valid, but the > > > contents during startup should always be 0. > > > > > > Can you do a "sysctl hw.busdma" while X is up with drm enabled? > > > > > > > > Here's the sysctl output: > > hw.busdma.total_bpages: 288 > > hw.busdma.zone0.total_bpages: 288 > > hw.busdma.zone0.free_bpages: 288 > > hw.busdma.zone0.reserved_bpages: 0 > > hw.busdma.zone0.active_bpages: 0 > > hw.busdma.zone0.total_bounced: 0 > > hw.busdma.zone0.total_deferred: 0 > > hw.busdma.zone0.lowaddr: 0xffffffff > > hw.busdma.zone0.alignment: 4096 > > Ok, it doesn't seem to be related to bounce pages... But it was worth a > shot... > > > It doesn't seem to change whether I'm in X with DRM disabled or enabled > > with the display corruption. Also, it took me a few starts of X to get > > it to even happen, as it seems there is occasionally a pci Out of memory > > (-12) error causing DRM to be disabled as well: > > The ENOMEM case should be returned if the kernel is not able to allocate > 32MB of contiguous memory for for the scatter gather pages backing the > GART. It technically doesn't need to be contiguous, but current methods > don't give us a choice. > > If I could figure out what bit of code to look at, it would help a > lot... Right now, I still don't see any issues in the drm code, so I'm > having to try and poke around in bus_dma and the vm to see if I can find > a bug there. We set the flags for those pages to be zeroed, but it > seems like it doesn't get flushed somehow. > > robert. > > > drmOpenDevice: node name is /dev/dri/card0 > > drmOpenDevice: open result is 10, (OK) > > drmOpenDevice: node name is /dev/dri/card0 > > drmOpenDevice: open result is 10, (OK) > > drmOpenByBusid: Searching for BusID pci:0000:04:00.0 > > drmOpenDevice: node name is /dev/dri/card0 > > drmOpenDevice: open result is 10, (OK) > > drmOpenByBusid: drmOpenMinor returns 10 > > drmOpenByBusid: drmGetBusid reports pci:0000:04:00.0 > > (II) [drm] DRM interface version 1.2 > > (II) [drm] DRM open master succeeded. > > (II) RADEON(0): [drm] Using the DRM lock SAREA also for drawables. > > (II) RADEON(0): [drm] framebuffer handle = 0xe8000000 > > (II) RADEON(0): [drm] added 1 reserved context for kernel > > (II) RADEON(0): X context handle = 0x1 > > (II) RADEON(0): [drm] installed DRM signal handler > > (EE) RADEON(0): [pci] Out of memory (-12) > > (EE) RADEON(0): [pci] PCI failed to initialize. Disabling the DRI. > > (II) RADEON(0): [drm] removed 1 reserved context for kernel > > (II) RADEON(0): [drm] unmapping 8192 bytes of SAREA 0xd06f8000 at 0x286fd000 > > (II) RADEON(0): [drm] Closed DRM master. > > (II) RADEON(0): RADEONRestoreMemMapRegisters() : > > (II) RADEON(0): MC_FB_LOCATION : 0xefffe000 0x1fff0000 > > (II) RADEON(0): MC_AGP_LOCATION : 0xffffffc0 > > (==) RADEON(0): Backing store disabled > > (WW) RADEON(0): Direct rendering disabled > > (II) RADEON(0): Render acceleration enabled for R200 type cards. > > (II) RADEON(0): Setting EXA maxPitchBytes > > > > Thanks for taking a look at this again. Could you try the attached patch and see if it makes any difference? My assembly foo is weak on i386. Jhb, Does this look at all correct? robert. > > _______________________________________________ > > freebsd-x11@freebsd.org mailing list > > http://lists.freebsd.org/mailman/listinfo/freebsd-x11 > > To unsubscribe, send any mail to "freebsd-x11-unsubscribe@freebsd.org" -- Robert Noland <rnoland@FreeBSD.org> FreeBSD --=-YktE7EncAO1fO72SX6x0 Content-Disposition: attachment; filename="pagezero.patch" Content-Type: text/x-patch; name="pagezero.patch"; charset="us-ascii" Content-Transfer-Encoding: 7bit diff --git a/sys/i386/i386/support.s b/sys/i386/i386/support.s index cbf0d82..49c5a63 100644 --- a/sys/i386/i386/support.s +++ b/sys/i386/i386/support.s @@ -372,7 +372,7 @@ ENTRY(sse2_pagezero) addl $4,%ecx cmpl %ecx,%eax jne 1b - sfence + mfence popl %ebx ret END(sse2_pagezero) --=-YktE7EncAO1fO72SX6x0--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1260556637.2281.19.camel>