On Wed, Aug 24, 2011 at 2:14 AM, Chen Jie <[email protected]> wrote: > Hi, > > 2011/8/23 Alex Deucher <[email protected]> >> >> > Not sure I understand it or not, but we're using the sideport way, and >> > has a dedicated 128M DDR2 memory as VRAM. >> >> The default behavior of the system bios is to set up sideport memory >> interleaved with stolen system memory. Unless your bios only enables >> sideport you'll need to respect the stolen system memory used as vram. >> Also, sideport memory has really limited memory bandwidth. It's a >> powersaving feature as if you un-interleave the sideport memory, you >> can put the display in sideport and stop memory access via the CPU. >> For decent performance, you need to use system memory or interleaved >> sideport and system memory. > Got it, thanks. > BTW, if using 128M sideport memory, will it occupy another 128M system > memory in the default behavior? >
Yes. Normally the sbios reports a slightly smaller ram size to account for the stolen memory or marks it as reserved so the OS doesn't try and use it. Alex > >> >> > >> > The details are at: >> > * >> > http://dev.lemote.com/files/upload/software/temp/Radeon.test/radeon_test.c >> > * diff with original radeon_test_moves(): >> > http://dev.lemote.com/files/upload/software/temp/Radeon.test/radeon_test.diff >> > >> > The new bo moving test is pass, but when I changed the bo size to 4M, >> > the kernel panic with unaligned access at some point after bo moving >> > test. The original radeon_test_moves() is ok for 4M bo size. Could >> > someone reviews the new radeon_test_moves() to find if this was caused >> > by some mistakes of the code or is a potential bug of the platform? >> >> See if the attached patch helps. It flushes the HDP caches if the >> driver uses the wait_idle callback rather than flushing in the fence >> emit. > The patch doesn't work for me. P.S. due to we haven't finished a > solid port to a newer version of kernel, the hack was done on the old > 2.6.36 kernel(http://dev.lemote.com/cgit/linux-3A.git/) > > I did some more hack, and found if create/reserve/pin more bo in VRAM > than available, it will cause unaligned access panic. And a simpler > way to trigger kernel panic(but another kind of kernel panic): > > struct radeon_bo **vram_obj = NULL; > uint64_t *vram_addr = NULL; > unsigned i, n, size; > > n = 31; /* 30 is the maximum num of bo can be allocated successfully */ > size = 4 * 1024 * 1024; > > vram_obj = kzalloc(n * sizeof(*vram_obj), GFP_KERNEL); > vram_addr = kzalloc(n * sizeof(*vram_addr), GFP_KERNEL); > > for (i = 0; i < n; i++) { > int r; > /* Create bo lives in VRAM */ > r = radeon_bo_create(rdev, NULL, size, true, RADEON_GEM_DOMAIN_VRAM, > vram_obj + i); > if (r) { > DRM_ERROR("Failed to create VRAM object %d\n", i); > break; > } > > r = radeon_bo_reserve(vram_obj[i], false); > if (unlikely(r != 0)) > break; > r = radeon_bo_pin(vram_obj[i], RADEON_GEM_DOMAIN_VRAM, vram_addr + i); > if (r) { > DRM_ERROR("Failed to pin VRAM object %d\n", i); > break; > } > } > > /* If I comment the following, no kernel panic for me */ > if (vram_obj) { > for (i = 0; i < n; i++) { > if (vram_obj[i]) { > if (radeon_bo_is_reserved(vram_obj[i])) { > radeon_bo_unpin(vram_obj[i]); > radeon_bo_unreserve(vram_obj[i]); > } > radeon_bo_unref(&vram_obj[i]); > } > } > kfree(vram_obj); > } > if (vram_addr) > kfree(vram_addr); > ---- > The corresponding dmesg log: > > [ 1.125000] [drm] Initialized drm 1.1.0 20060810 > [ 1.128906] [drm] radeon defaulting to kernel modesetting. > [ 1.136718] [drm] radeon kernel modesetting enabled. > [ 1.175781] [drm] initializing kernel modesetting (RS780 0x1002:0x9615). > [ 1.183593] [drm] register mmio base: 0x48000000 > [ 1.187500] [drm] register mmio size: 65536 > [ 1.195312] radeon 0000:01:05.0: BAR 6: can't assign [??? > 0x00000000 flags 0x0] (bogus alignment) > [ 1.203125] ATOM BIOS: BR041389 > [ 1.207031] radeon 0000:01:05.0: VRAM: 128M 0x10000000 - 0x17FFFFFF > (128M used) > [ 1.214843] radeon 0000:01:05.0: GTT: 512M 0x18000000 - 0x37FFFFFF > [ 1.222656] [drm] Detected VRAM RAM=128M, BAR=128M > [ 1.226562] [drm] RAM width 32bits DDR > [ 1.230468] [TTM] Zone kernel: Available graphics memory: 464720 kiB. > [ 1.238281] [TTM] Initializing pool allocator. > [ 1.242187] [drm] radeon: 128M of VRAM memory ready > [ 1.246093] [drm] radeon: 512M of GTT memory ready. > [ 1.250000] [drm] radeon: irq initialized. > [ 1.253906] [drm] GART: num cpu pages 32768, num gpu pages 131072 > [ 1.265625] [drm] Loading RS780 Microcode > [ 1.343750] [drm] ring test succeeded in 0 usecs > [ 1.347656] [drm] radeon: ib pool ready. > [ 1.351562] [drm] ib test succeeded in 0 usecs > [ 1.355468] [drm] Enabling audio support > [ 1.359375] radeon 0000:01:05.0: 980000013eecbe00 pin failed > [ 1.367187] [drm:radeon_test_moves] *ERROR* Failed to pin VRAM object 31 > [ 1.378906] radeon 0000:01:05.0: 980000013eecbe00 unpin not necessary > [ 1.386718] CPU 3 Unable to handle kernel paging request at virtual > address 0000000000000008, epc == ffffffff804aff08, ra == > ffffffff804aff00 > [ 1.398437] Oops[#1]: > [ 1.402343] Cpu 3 > [ 1.402343] $ 0 : 0000000000000000 000000001400fce0 > 980000013eecbe00 0000000000000000 > [ 1.402343] $ 4 : 980000013e5e2f50 ffffffff8047a66c > 0000000000000000 0000000000000000 > [ 1.402343] $ 8 : 0000000000000000 980000000161c000 > 00004124924d2492 ffffffff81073f20 > [ 1.402343] $12 : 0000000000000007 ffffffff810f3ee0 > fffffffffffffffc ffffffff81073f28 > [ 1.402343] $16 : 980000013eecbe00 980000013eecbe48 > 980000013e8f7d40 000000000000001f > [ 1.402343] $20 : 980000013e8f7e00 980000013e8f7ef8 > 0000000000000028 980000013e5e2000 > [ 1.402343] $24 : 0000000000000000 ffffffff804afedc > [ 1.402343] $28 : 980000013e044000 980000013e047af0 > 980000013eecbe00 ffffffff804aff00 > [ 1.402343] Hi : 0000000000000000 > [ 1.402343] Lo : 00000000011694a8 > [ 1.402343] epc : ffffffff804aff08 radeon_ttm_bo_destroy+0x2c/0x68 > [ 1.402343] Not tainted > [ 1.402343] ra : ffffffff804aff00 radeon_ttm_bo_destroy+0x24/0x68 > [ 1.402343] Status: 1400fce3 KX SX UX KERNEL EXL IE > [ 1.402343] Cause : 0000000c > [ 1.402343] BadVA : 0000000000000008 > [ 1.402343] PrId : 00006305 (ICT Loongson-3) > [ 1.402343] Modules linked in: > [ 1.402343] Process swapper (pid: 1, threadinfo=980000013e044000, > task=980000013e078000, tls=0000000000000000) > [ 1.402343] Stack : 980000013eecbe8c 980000013eecbe8c > ffffffff8047a66c ffffffff803f056c > [ 1.402343] 0000000000000028 980000013eecbe48 > 980000013e5e2530 ffffffff8047b478 > [ 1.402343] 0000000000070000 980000013eecbe88 > ffffffff8047b418 ffffffff803f056c > [ 1.402343] 000000000000001f 980000013e5e2530 > 980000013eecbe48 ffffffff8047a950 > [ 1.402343] 000000001400fce1 980000013e8f7cf8 > 980000013e5e3dc0 ffffffff804b0230 > [ 1.402343] 0000000000000000 ffffffff8047a554 > 980000013e8f7cf8 980000013e8f7cf8 > [ 1.402343] 980000013e8f7c00 ffffffff804f27bc > 980000013e5e2000 0000000000000000 > [ 1.402343] 00000000ffffffff 0000000000c20021 > 980000013e5db800 980000013e5dbbb8 > [ 1.402343] 0000000000000004 0000000000000000 > 0000000000000000 ffffffff8049b85c > [ 1.402343] 980000013e5db800 980000013e5e2000 > fffffffffffffff4 ffffffff8049cc08 > [ 1.402343] ... > [ 1.402343] Call Trace: > [ 1.402343] [<ffffffff804aff08>] radeon_ttm_bo_destroy+0x2c/0x68 > [ 1.402343] [<ffffffff803f056c>] kref_put+0x70/0x90 > [ 1.402343] [<ffffffff8047b478>] ttm_bo_release+0x60/0x78 > [ 1.402343] [<ffffffff803f056c>] kref_put+0x70/0x90 > [ 1.402343] [<ffffffff8047a950>] ttm_bo_unref+0x38/0x50 > [ 1.402343] [<ffffffff804b0230>] radeon_bo_unref+0x3c/0x68 > [ 1.402343] [<ffffffff804f27bc>] radeon_test_moves+0x174/0x1d0 > [ 1.402343] [<ffffffff8049b85c>] radeon_device_init+0x474/0x4b4 > [ 1.402343] [<ffffffff8049cc08>] radeon_driver_load_kms+0xcc/0x138 > [ 1.402343] [<ffffffff8046c54c>] drm_get_pci_dev+0x140/0x230 > [ 1.402343] [<ffffffff80406af0>] local_pci_probe+0x5c/0xd8 > [ 1.402343] [<ffffffff80406f48>] pci_device_probe+0x5c/0x9c > [ 1.402343] [<ffffffff80528690>] driver_probe_device+0xe4/0x1c0 > [ 1.402343] [<ffffffff805287d8>] __driver_attach+0x6c/0xa4 > [ 1.402343] [<ffffffff8052764c>] bus_for_each_dev+0x54/0x94 > [ 1.402343] [<ffffffff80527e48>] bus_add_driver+0xbc/0x258 > [ 1.402343] [<ffffffff80528ae4>] driver_register+0xc4/0x178 > [ 1.402343] [<ffffffff80407310>] __pci_register_driver+0x50/0xe0 > [ 1.402343] [<ffffffff802159dc>] do_one_initcall+0x88/0x174 > [ 1.402343] [<ffffffff81078c60>] kernel_init+0x208/0x2bc > [ 1.402343] [<ffffffff80217c84>] kernel_thread_helper+0x10/0x18 > [ 1.402343] > [ 1.402343] > [ 1.402343] Code: 64840f50 de020008 de23ffb8 <fc620008> fc430000 > de0401e0 fe30ffb8 fe100008 0c084d49 > [ 1.402343] Disabling lock debugging due to kernel taint > [ 1.750000] Kernel panic - not syncing: Attempted to kill init! > > > > Regards, > - Chen Jie > _______________________________________________ xorg-driver-ati mailing list [email protected] http://lists.x.org/mailman/listinfo/xorg-driver-ati
