dri-devel June 2012

dri-devel@lists.freedesktop.org

125 participants
696 discussions

[Bug 17902] [830M missing dvo driver] need support for DVO-LVDS via na2501
by bugzilla-daemon＠freedesktop.org 07 Jun '12

07 Jun '12

https://bugs.freedesktop.org/show_bug.cgi?id=17902 --- Comment #84 from Daniel Vetter <daniel(a)ffwll.ch> 2012-06-07 00:32:02 PDT --- On Thu, Jun 7, 2012 at 3:21 AM, <bugzilla-daemon(a)freedesktop.org> wrote: > (Is actually anyone reading this??? Is there a better way to supply patches?) Well, if you have a patch that works somewhat attaching it to all the relevant bugzillas is usually a good way to get test feedback. If the entire thing works reasonably well (which for this old hw equals 'it works for you somewhat', no high standards), please submit the complete kernel patch with a nice changelog to the intel-gfx mailing list so that it can be merged. -Daniel -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.

1 0

[Bug 17902] [830M missing dvo driver] need support for DVO-LVDS via na2501
by bugzilla-daemon＠freedesktop.org 07 Jun '12

07 Jun '12

https://bugs.freedesktop.org/show_bug.cgi?id=17902 --- Comment #83 from thor(a)math.tu-berlin.de 2012-06-06 18:21:06 PDT --- Finally, success! I'm not quite sure why, but for reasons unclear to me the DVO chip only wants to talk if the PLL is enabled and running and the screen resolution fits. In addition, the system has apparently a "fake" VGA output that must be disabled to have the system working properly. Otherwise, the KMS layer seems to want to redirect the output to the VGA, and then drops dead, leaving an unusable screen (and DVO!) behind. I do not know yet whether the physical VGA connector works, but basically, here is the receipt: 1) Apply the patches I mentioned above. Especially, the bypass mode of the scaler must be set as its function is unclear. static void ns2501_dpms(struct intel_dvo_device *dvo, int mode) { struct ns2501_priv *ns = (struct ns2501_priv *)(dvo->dev_priv); unsigned char ch; DRM_DEBUG_KMS("%s: Trying set the dpms of the DVO to %d\n",__FUNCTION__,mode); if (ns->reg_8_set) { ch = ns->reg_8_shadow; } else { ch = NS2501_8_PD | NS2501_8_BPAS | NS2501_8_VEN | NS2501_8_HEN; } if (mode == DRM_MODE_DPMS_ON) ch |= NS2501_8_PD; else ch &= ~NS2501_8_PD; ch |= NS2501_8_BPAS; if (ns->reg_8_set == 0 || ns->reg_8_shadow != ch) { ns->reg_8_set = 1; ns->reg_8_shadow = ch; ns2501_writeb(dvo, NS2501_REG8, ch); } } Here I added a shadow register which is likely not exactly necessary, though I wanted to avoid the i2c communication if not necessary. 2) There is still a bug in the ns2501 source in so far as the query function returns something unitialized if reading from the DVO fails. And it fails quite often. static enum drm_connector_status ns2501_detect(struct intel_dvo_device *dvo) { uint8_t reg9; enum drm_connector_status status = connector_status_unknown; struct ns2501_priv *ns = (struct ns2501_priv *)(dvo->dev_priv); DRM_DEBUG_KMS("%s: Trying to detect the connector status\n",__FUNCTION__); if (ns2501_readb(dvo, NS2501_REG9, &reg9)) { if (!(reg9 & NS2501_9_RSEN)) status = connector_status_connected; else status = connector_status_disconnected; ns->reg_9_set = 1; ns->reg_9_shadow = reg9; } else if (ns->reg_9_set) { if (!(ns->reg_9_shadow & NS2501_9_RSEN)) status = connector_status_connected; else status = connector_status_disconnected; } DRM_DEBUG_KMS("%s: Status is %d\n",__FUNCTION__,status); return status; } Again a shadow register was added. 3) The following kernel arguments should be added to disable the VGA output: i915.modeset=1 video=VGA-1:1024x768d video=DVI-I-1:1024x768e I'm not sure why the system behaives so strange otherwise, I'll try to dig a bit deeper, but I believe that there is some limitation with the i830 and its outputs. Probably I'll find a datasheet. With the above modifications, you cannot, of course, scale the screen (resolution is fixed), though 3D acceleration works nicely. The same trick might also apply to the IBM R31, I'll check later next week - it had similar issues. (Is actually anyone reading this??? Is there a better way to supply patches?) -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.

1 0

radeon gpu driver bug on suspend/resume in 3.5-rc1
by Austin Lund 07 Jun '12

07 Jun '12

Hi, I get the attached traces with 3.5-rc1 after suspend/resume, sometimes. It doesn't always happen. Usually happens at least once in 10 suspend/resume cycles. The first trace seems non fatal, but the system locks up in the second one and needs to be rebooted. =========================================================================== [ 1435.785548] Restarting tasks ... done. [ 1435.854748] usb 2-1.2: new full-speed USB device number 5 using ehci_hcd [ 1435.856616] [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -35! [ 1435.857653] radeon 0000:01:00.0: GPU reset succeed [ 1435.878746] [drm] PCIE GART of 512M enabled (table at 0x0000000000040000). [ 1435.878837] radeon 0000:01:00.0: WB enabled [ 1435.878839] radeon 0000:01:00.0: fence driver on ring 0 use gpu addr 0x0000000040000c00 and cpu addr 0xffff88025e8c2c00 [ 1435.895010] [drm] ring test on 0 succeeded in 2 usecs [ 1435.895067] [drm] ib test on ring 0 succeeded in 0 usecs =========================================================================== [ 1376.089570] Restarting tasks ... done. [ 1376.092574] radeon 0000:01:00.0: GPU lockup CP stall for more than 17300msec [ 1376.092615] radeon 0000:01:00.0: GPU lockup (waiting for 0x0000000000019986 last fence id 0x0000000000019981) [ 1376.093654] radeon 0000:01:00.0: GPU softreset [ 1376.093657] radeon 0000:01:00.0: GRBM_STATUS=0xA0003828 [ 1376.093659] radeon 0000:01:00.0: GRBM_STATUS_SE0=0x00000007 [ 1376.093662] radeon 0000:01:00.0: GRBM_STATUS_SE1=0x00000007 [ 1376.093664] radeon 0000:01:00.0: SRBM_STATUS=0x200000C0 [ 1376.093675] radeon 0000:01:00.0: GRBM_SOFT_RESET=0x00007F6B [ 1376.093778] radeon 0000:01:00.0: GRBM_STATUS=0x00003828 [ 1376.093781] radeon 0000:01:00.0: GRBM_STATUS_SE0=0x00000007 [ 1376.093783] radeon 0000:01:00.0: GRBM_STATUS_SE1=0x00000007 [ 1376.093785] radeon 0000:01:00.0: SRBM_STATUS=0x200000C0 [ 1376.094781] radeon 0000:01:00.0: GPU reset succeed [ 1376.116071] [drm] PCIE GART of 512M enabled (table at 0x0000000000040000). [ 1376.116165] radeon 0000:01:00.0: WB enabled [ 1376.116167] radeon 0000:01:00.0: fence driver on ring 0 use gpu addr 0x0000000040000c00 and cpu addr 0xffff88025e42ec00 [ 1376.132330] [drm] ring test on 0 succeeded in 2 usecs [ 1376.132392] [drm] ib test on ring 0 succeeded in 0 usecs [ 1376.741682] tg3 0000:02:00.0: irq 49 for MSI/MSI-X [ 1376.741703] tg3 0000:02:00.0: irq 50 for MSI/MSI-X [ 1376.741718] tg3 0000:02:00.0: irq 51 for MSI/MSI-X [ 1376.741733] tg3 0000:02:00.0: irq 52 for MSI/MSI-X [ 1376.741748] tg3 0000:02:00.0: irq 53 for MSI/MSI-X [ 1377.772057] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready [ 1377.926958] b43-phy0: Loading firmware version 666.2 (2011-02-23 01:15:07) [ 1377.966914] b43-phy0 debug: Chip initialized [ 1377.967124] b43-phy0 debug: 64-bit DMA initialized [ 1377.967199] b43-phy0 debug: QoS enabled [ 1377.968068] b43-phy0 debug: Wireless interface started [ 1377.968140] b43-phy0 debug: Adding Interface type 2 [ 1377.969962] IPv6: ADDRCONF(NETDEV_UP): wlan0: link is not ready [ 1381.080635] wlan0: authenticate with 00:12:bf:12:7a:49 [ 1381.092133] wlan0: direct probe to 00:12:bf:12:7a:49 (try 1/3) [ 1381.295524] wlan0: direct probe to 00:12:bf:12:7a:49 (try 2/3) [ 1381.499372] wlan0: direct probe to 00:12:bf:12:7a:49 (try 3/3) [ 1381.703233] wlan0: authentication with 00:12:bf:12:7a:49 timed out [ 1386.914207] radeon 0000:01:00.0: GPU lockup CP stall for more than 10000msec [ 1386.914218] radeon 0000:01:00.0: GPU lockup (waiting for 0x000000000001998b last fence id 0x0000000000019988) [ 1386.914225] radeon 0000:01:00.0: ffff8802605ec000 pin failed [ 1386.986828] radeon 0000:01:00.0: couldn't schedule ib [ 1386.986837] [drm:radeon_cs_ib_chunk] *ERROR* Failed to schedule IB ! [ 1386.986856] ------------[ cut here ]------------ [ 1386.986899] WARNING: at /home/lund/src/linux/drivers/gpu/drm/radeon/radeon_fence.c:180 radeon_fence_signaled+0x8d/0x90 [radeon]() [ 1386.986902] Hardware name: MacBookPro8,2 [ 1386.986905] Querying an unemitted fence : ffff8802670045a0 ! [ 1386.986972] Modules linked in: autofs4 snd_hda_codec_hdmi rfcomm bnep snd_hda_codec_cirrus arc4 snd_hda_intel b43 snd_hda_codec ghash_clmulni_intel aesni_intel radeon snd_hwdep snd_pcm snd_seq_midi cryptd binfmt_misc snd_rawmidi mac80211 aes_x86_64 ttm microcode drm_kms_helper drm cfg80211 uvcvideo videobuf2_core ssb applesmc videodev videobuf2_vmalloc input_polldev i2c_algo_bit btusb apple_gmux videobuf2_memops hid_generic video bluetooth bcma snd_seq_midi_event bcm5974 snd_seq snd_timer snd_seq_device snd soundcore snd_page_alloc apple_bl hid_apple firewire_ohci firewire_core usbhid tg3 hid crc_itu_t [ 1386.986980] Pid: 1338, comm: Xorg Not tainted 3.5.0-rc1 #21 [ 1386.986983] Call Trace: [ 1386.986998] [<ffffffff8103a3cf>] warn_slowpath_common+0x7f/0xc0 [ 1386.987006] [<ffffffff8103a4c6>] warn_slowpath_fmt+0x46/0x50 [ 1386.987034] [<ffffffffa0360d3d>] radeon_fence_signaled+0x8d/0x90 [radeon] [ 1386.987060] [<ffffffffa036190e>] radeon_sync_obj_signaled+0xe/0x10 [radeon] [ 1386.987073] [<ffffffffa02618f0>] ttm_bo_wait+0x130/0x1b0 [ttm] [ 1386.987085] [<ffffffffa0262acb>] ? ttm_bo_reserve+0x7b/0xa0 [ttm] [ 1386.987113] [<ffffffffa0363f69>] radeon_bo_wait+0x89/0xc0 [radeon] [ 1386.987147] [<ffffffffa0375ab5>] radeon_gem_wait_idle_ioctl+0x45/0xd0 [radeon] [ 1386.987167] [<ffffffffa01e64cc>] drm_ioctl+0x47c/0x540 [drm] [ 1386.987175] [<ffffffff8117afa4>] ? mntput+0x24/0x40 [ 1386.987206] [<ffffffffa0375a70>] ? radeon_gem_busy_ioctl+0x100/0x100 [radeon] [ 1386.987214] [<ffffffff8119afb2>] ? fsnotify+0x1d2/0x2b0 [ 1386.987222] [<ffffffff8116f178>] do_vfs_ioctl+0x98/0x550 [ 1386.987229] [<ffffffff8127eb97>] ? apparmor_capable+0x27/0xa0 [ 1386.987236] [<ffffffff8116f6c1>] sys_ioctl+0x91/0xa0 [ 1386.987245] [<ffffffff8150317e>] ? do_device_not_available+0xe/0x10 [ 1386.987252] [<ffffffff8150a112>] system_call_fastpath+0x16/0x1b [ 1386.987257] ---[ end trace ff51fe717d675026 ]--- [ 1387.041174] radeon 0000:01:00.0: couldn't schedule ib [ 1387.041180] [drm:radeon_cs_ib_chunk] *ERROR* Failed to schedule IB ! [ 1387.041213] ------------[ cut here ]------------ [ 1387.041247] WARNING: at /home/lund/src/linux/drivers/gpu/drm/radeon/radeon_fence.c:180 radeon_fence_signaled+0x8d/0x90 [radeon]() [ 1387.041250] Hardware name: MacBookPro8,2 [ 1387.041253] Querying an unemitted fence : ffff8802670045a0 ! [ 1387.041315] Modules linked in: autofs4 snd_hda_codec_hdmi rfcomm bnep snd_hda_codec_cirrus arc4 snd_hda_intel b43 snd_hda_codec ghash_clmulni_intel aesni_intel radeon snd_hwdep snd_pcm snd_seq_midi cryptd binfmt_misc snd_rawmidi mac80211 aes_x86_64 ttm microcode drm_kms_helper drm cfg80211 uvcvideo videobuf2_core ssb applesmc videodev videobuf2_vmalloc input_polldev i2c_algo_bit btusb apple_gmux videobuf2_memops hid_generic video bluetooth bcma snd_seq_midi_event bcm5974 snd_seq snd_timer snd_seq_device snd soundcore snd_page_alloc apple_bl hid_apple firewire_ohci firewire_core usbhid tg3 hid crc_itu_t [ 1387.041322] Pid: 1338, comm: Xorg Tainted: G W 3.5.0-rc1 #21 [ 1387.041325] Call Trace: [ 1387.041337] [<ffffffff8103a3cf>] warn_slowpath_common+0x7f/0xc0 [ 1387.041345] [<ffffffff8103a4c6>] warn_slowpath_fmt+0x46/0x50 [ 1387.041372] [<ffffffffa0360d3d>] radeon_fence_signaled+0x8d/0x90 [radeon] [ 1387.041399] [<ffffffffa036190e>] radeon_sync_obj_signaled+0xe/0x10 [radeon] [ 1387.041411] [<ffffffffa02618f0>] ttm_bo_wait+0x130/0x1b0 [ttm] [ 1387.041423] [<ffffffffa0262acb>] ? ttm_bo_reserve+0x7b/0xa0 [ttm] [ 1387.041452] [<ffffffffa0363f69>] radeon_bo_wait+0x89/0xc0 [radeon] [ 1387.041486] [<ffffffffa03759c0>] radeon_gem_busy_ioctl+0x50/0x100 [radeon] [ 1387.041505] [<ffffffffa01e64cc>] drm_ioctl+0x47c/0x540 [drm] [ 1387.041511] [<ffffffff8117afa4>] ? mntput+0x24/0x40 [ 1387.041543] [<ffffffffa0375970>] ? radeon_gem_mmap_ioctl+0x20/0x20 [radeon] [ 1387.041551] [<ffffffff8119afb2>] ? fsnotify+0x1d2/0x2b0 [ 1387.041558] [<ffffffff8116f178>] do_vfs_ioctl+0x98/0x550 [ 1387.041565] [<ffffffff8127eb97>] ? apparmor_capable+0x27/0xa0 [ 1387.041571] [<ffffffff8116f6c1>] sys_ioctl+0x91/0xa0 [ 1387.041579] [<ffffffff8150317e>] ? do_device_not_available+0xe/0x10 [ 1387.041586] [<ffffffff8150a112>] system_call_fastpath+0x16/0x1b [ 1387.041590] ---[ end trace ff51fe717d675027 ]---

1 0

Re: Circular locking (and possible deadlock), when exiting from mplayer -vo fbdev2
by Dan Carpenter 06 Jun '12

06 Jun '12

Hi Dri devs, Witold Baryluk <baryluk(a)smp.if.uj.edu.pl> reported a lockdep splat to the fbdev list, but apparently that was the wrong list and you people are the right list to handle this. Here is the lockdep splat and Witold's config: http://marc.info/?l=linux-fbdev&m=133883191129462&w=2 I looked into the bug and it turns out there is a lock ordering problem but I'm not sure how to fix it. http://marc.info/?l=linux-fbdev&m=133890563322819&w=2 Cut and pasted from that URL: ----------- The problem is we hold ->mmap_sem when we call fb_release() which takes the info->lock. We take the info->lock in do_fb_ioctl() before we call fb_set_var() which calls drm_fb_helper_set_par() which takes the mode_config.mutex. In drm_mode_getresources() we take the mode_config.mutex() and call put_user() which takes the ->mmap_sem. So on one CPU we are holding the ->mmap_sem and want the info->lock. On another CPU we are holding the info->lock and want the config.mutex. On the other CPU we hold the config.mutex and want the ->mmap_sem. Deadlock. ----------- Could you take a look? regards, dan carpenter

1 0

[PATCH 1/3] intel: sanitize i915_drm.h
by Ben Widawsky 06 Jun '12

06 Jun '12

run make headers_isntall on d-i-n, copy to here Signed-off-by: Ben Widawsky <ben(a)bwidawsk.net> --- include/drm/i915_drm.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h index af3ce17..725a8de 100644 --- a/include/drm/i915_drm.h +++ b/include/drm/i915_drm.h @@ -33,6 +33,7 @@ * subject to backwards-compatibility constraints. */ + /* Each region is a minimum of 16k, and there are at most 255 of them. */ #define I915_NR_TEX_REGIONS 255 /* table size 2k - maximum due to use @@ -287,7 +288,8 @@ typedef struct drm_i915_irq_wait { #define I915_PARAM_HAS_EXEC_CONSTANTS 14 #define I915_PARAM_HAS_RELAXED_DELTA 15 #define I915_PARAM_HAS_GEN7_SOL_RESET 16 -#define I915_PARAM_HAS_LLC 17 +#define I915_PARAM_HAS_LLC 17 +#define I915_PARAM_HAS_ALIASING_PPGTT 18 typedef struct drm_i915_getparam { int param; -- 1.7.10.3

2 9

[PATCH] drm/radeon/kms: attempt to fix rs690 issues.
by Dave Airlie 06 Jun '12

06 Jun '12

From: Dave Airlie <airlied(a)redhat.com> Since c9a1be96277b3b2d2e8aff2ba69d7817ea8e46c9, or at least since v3.2 we've had some regression reports, this is an attempt to fix them by putting back some volatiles that were removed in that commit. Signed-off-by: Dave Airlie <airlied(a)redhat.com> --- drivers/gpu/drm/radeon/r100.c | 2 +- drivers/gpu/drm/radeon/r300.c | 2 +- drivers/gpu/drm/radeon/radeon.h | 2 +- drivers/gpu/drm/radeon/radeon_gart.c | 2 +- drivers/gpu/drm/radeon/rs400.c | 2 +- 5 files changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index fb44e7e..26488f7 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -680,7 +680,7 @@ void r100_pci_gart_disable(struct radeon_device *rdev) int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) { - u32 *gtt = rdev->gart.ptr; + volatile u32 *gtt = rdev->gart.ptr; if (i < 0 || i > rdev->gart.num_gpu_pages) { return -EINVAL; diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index 97722a3..bb596f4 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -74,7 +74,7 @@ void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev) int rv370_pcie_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) { - void __iomem *ptr = rdev->gart.ptr; + void __iomem *ptr = (void *)rdev->gart.ptr; if (i < 0 || i > rdev->gart.num_gpu_pages) { return -EINVAL; diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index fefcca5..3fd759a 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -470,7 +470,7 @@ struct radeon_mc; struct radeon_gart { dma_addr_t table_addr; struct radeon_bo *robj; - void *ptr; + volatile void *ptr; unsigned num_gpu_pages; unsigned num_cpu_pages; unsigned table_size; diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c index 59d4493..b2db83a 100644 --- a/drivers/gpu/drm/radeon/radeon_gart.c +++ b/drivers/gpu/drm/radeon/radeon_gart.c @@ -102,7 +102,7 @@ int radeon_gart_table_vram_pin(struct radeon_device *rdev) radeon_bo_unreserve(rdev->gart.robj); return r; } - r = radeon_bo_kmap(rdev->gart.robj, &rdev->gart.ptr); + r = radeon_bo_kmap(rdev->gart.robj, (void **)&rdev->gart.ptr); if (r) radeon_bo_unpin(rdev->gart.robj); radeon_bo_unreserve(rdev->gart.robj); diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c index a464eb5..82f1a7c 100644 --- a/drivers/gpu/drm/radeon/rs400.c +++ b/drivers/gpu/drm/radeon/rs400.c @@ -212,7 +212,7 @@ void rs400_gart_fini(struct radeon_device *rdev) int rs400_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) { uint32_t entry; - u32 *gtt = rdev->gart.ptr; + volatile u32 *gtt = rdev->gart.ptr; if (i < 0 || i > rdev->gart.num_gpu_pages) { return -EINVAL; -- 1.7.10.2

1 0

[PATCH] drm: Remove unused fields from drm_display_mode
by Adam Jackson 06 Jun '12

06 Jun '12

% size drivers/gpu/drm/drm.o* text data bss dec hex filename 187371 8299 336 196006 2fda6 drivers/gpu/drm/drm.o 191251 8299 336 199886 30cce drivers/gpu/drm/drm.o.orig Signed-off-by: Adam Jackson <ajax(a)redhat.com> --- drivers/gpu/drm/drm_crtc.c | 3 +-- drivers/gpu/drm/drm_crtc_helper.c | 2 -- drivers/gpu/drm/drm_edid.c | 2 +- drivers/gpu/drm/drm_fb_helper.c | 2 +- drivers/gpu/drm/drm_modes.c | 16 +++------------- drivers/gpu/drm/exynos/exynos_drm_connector.c | 1 - drivers/gpu/drm/exynos/exynos_drm_crtc.c | 2 +- drivers/gpu/drm/exynos/exynos_hdmi.c | 4 ++-- drivers/gpu/drm/gma500/psb_drv.c | 1 - drivers/gpu/drm/i2c/ch7006_mode.c | 1 - drivers/gpu/drm/nouveau/nouveau_connector.c | 5 ++--- drivers/gpu/drm/nouveau/nv04_crtc.c | 3 +++ drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 2 -- include/drm/drm_crtc.h | 10 +--------- 14 files changed, 15 insertions(+), 39 deletions(-) diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 08a7aa7..f367dc9 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -1080,7 +1080,7 @@ static void drm_crtc_convert_to_umode(struct drm_mode_modeinfo *out, out->vsync_end = in->vsync_end; out->vtotal = in->vtotal; out->vscan = in->vscan; - out->vrefresh = in->vrefresh; + out->vrefresh = drm_mode_vrefresh(in); out->flags = in->flags; out->type = in->type; strncpy(out->name, in->name, DRM_DISPLAY_MODE_LEN); @@ -1118,7 +1118,6 @@ static int drm_crtc_convert_umode(struct drm_display_mode *out, out->vsync_end = in->vsync_end; out->vtotal = in->vtotal; out->vscan = in->vscan; - out->vrefresh = in->vrefresh; out->flags = in->flags; out->type = in->type; strncpy(out->name, in->name, DRM_DISPLAY_MODE_LEN); diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c index 3252e70..089a838 100644 --- a/drivers/gpu/drm/drm_crtc_helper.c +++ b/drivers/gpu/drm/drm_crtc_helper.c @@ -159,8 +159,6 @@ prune: DRM_DEBUG_KMS("[CONNECTOR:%d:%s] probed modes :\n", connector->base.id, drm_get_connector_name(connector)); list_for_each_entry(mode, &connector->modes, head) { - mode->vrefresh = drm_mode_vrefresh(mode); - drm_mode_set_crtcinfo(mode, CRTC_INTERLACE_HALVE_V); drm_mode_debug_printmodeline(mode); } diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 5ab377f..0858757 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -450,7 +450,7 @@ static u32 edid_get_quirks(struct edid *edid) } #define MODE_SIZE(m) ((m)->hdisplay * (m)->vdisplay) -#define MODE_REFRESH_DIFF(m,r) (abs((m)->vrefresh - target_refresh)) +#define MODE_REFRESH_DIFF(m,r) (abs(drm_mode_vrefresh((m)) - target_refresh)) /** * edid_fixup_preferred - set preferred modes based on quirk list diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 5683b7f..4bee062 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -980,7 +980,7 @@ static struct drm_display_mode *drm_pick_cmdline_mode(struct drm_fb_helper_conne continue; if (cmdline_mode->refresh_specified) { - if (mode->vrefresh != cmdline_mode->refresh) + if (drm_mode_vrefresh(mode) != cmdline_mode->refresh) continue; } diff --git a/drivers/gpu/drm/drm_modes.c b/drivers/gpu/drm/drm_modes.c index 8df4ec8..f076062 100644 --- a/drivers/gpu/drm/drm_modes.c +++ b/drivers/gpu/drm/drm_modes.c @@ -49,9 +49,9 @@ */ void drm_mode_debug_printmodeline(struct drm_display_mode *mode) { - DRM_DEBUG_KMS("Modeline %d:\"%s\" %d %d %d %d %d %d %d %d %d %d " + DRM_DEBUG_KMS("Modeline %d:\"%s\" %d %d %d %d %d %d %d %d %d " "0x%x 0x%x\n", - mode->base.id, mode->name, mode->vrefresh, mode->clock, + mode->base.id, mode->name, mode->clock, mode->hdisplay, mode->hsync_start, mode->hsync_end, mode->htotal, mode->vdisplay, mode->vsync_start, @@ -598,9 +598,6 @@ int drm_mode_hsync(const struct drm_display_mode *mode) { unsigned int calc_val; - if (mode->hsync) - return mode->hsync; - if (mode->htotal < 0) return 0; @@ -621,8 +618,6 @@ EXPORT_SYMBOL(drm_mode_hsync); * * Return @mode's vrefresh rate in Hz or calculate it if necessary. * - * FIXME: why is this needed? shouldn't vrefresh be set already? - * * RETURNS: * Vertical refresh rate. It will be the result of actual value plus 0.5. * If it is 70.288, it will return 70Hz. @@ -633,9 +628,7 @@ int drm_mode_vrefresh(const struct drm_display_mode *mode) int refresh = 0; unsigned int calc_val; - if (mode->vrefresh > 0) - refresh = mode->vrefresh; - else if (mode->htotal > 0 && mode->vtotal > 0) { + if (mode->htotal > 0 && mode->vtotal > 0) { int vtotal; vtotal = mode->vtotal; /* work out vrefresh the value will be x1000 */ @@ -706,9 +699,6 @@ void drm_mode_set_crtcinfo(struct drm_display_mode *p, int adjust_flags) p->crtc_vblank_end = max(p->crtc_vsync_end, p->crtc_vtotal); p->crtc_hblank_start = min(p->crtc_hsync_start, p->crtc_hdisplay); p->crtc_hblank_end = max(p->crtc_hsync_end, p->crtc_htotal); - - p->crtc_hadjusted = false; - p->crtc_vadjusted = false; } EXPORT_SYMBOL(drm_mode_set_crtcinfo); diff --git a/drivers/gpu/drm/exynos/exynos_drm_connector.c b/drivers/gpu/drm/exynos/exynos_drm_connector.c index bf791fa..e841087 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_connector.c +++ b/drivers/gpu/drm/exynos/exynos_drm_connector.c @@ -51,7 +51,6 @@ convert_to_display_mode(struct drm_display_mode *mode, DRM_DEBUG_KMS("%s\n", __FILE__); mode->clock = timing->pixclock / 1000; - mode->vrefresh = timing->refresh; mode->hdisplay = timing->xres; mode->hsync_start = mode->hdisplay + timing->right_margin; diff --git a/drivers/gpu/drm/exynos/exynos_drm_crtc.c b/drivers/gpu/drm/exynos/exynos_drm_crtc.c index 4afb625..7289cf7 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_crtc.c +++ b/drivers/gpu/drm/exynos/exynos_drm_crtc.c @@ -120,7 +120,7 @@ int exynos_drm_overlay_update(struct exynos_drm_overlay *overlay, /* set drm mode data. */ overlay->mode_width = mode->hdisplay; overlay->mode_height = mode->vdisplay; - overlay->refresh = mode->vrefresh; + overlay->refresh = drm_mode_vrefresh(mode); overlay->scan_flag = mode->flags; DRM_DEBUG_KMS("overlay : offset_x/y(%d,%d), width/height(%d,%d)", diff --git a/drivers/gpu/drm/exynos/exynos_hdmi.c b/drivers/gpu/drm/exynos/exynos_hdmi.c index a137e9e..293c37d 100644 --- a/drivers/gpu/drm/exynos/exynos_hdmi.c +++ b/drivers/gpu/drm/exynos/exynos_hdmi.c @@ -1223,7 +1223,7 @@ static int hdmi_v13_conf_index(struct drm_display_mode *mode) for (i = 0; i < ARRAY_SIZE(hdmi_v13_confs); ++i) if (hdmi_v13_confs[i].width == mode->hdisplay && hdmi_v13_confs[i].height == mode->vdisplay && - hdmi_v13_confs[i].vrefresh == mode->vrefresh && + hdmi_v13_confs[i].vrefresh == drm_mode_vrefresh(mode) && hdmi_v13_confs[i].interlace == ((mode->flags & DRM_MODE_FLAG_INTERLACE) ? true : false)) @@ -1239,7 +1239,7 @@ static int hdmi_v14_conf_index(struct drm_display_mode *mode) for (i = 0; i < ARRAY_SIZE(hdmi_confs); ++i) if (hdmi_confs[i].width == mode->hdisplay && hdmi_confs[i].height == mode->vdisplay && - hdmi_confs[i].vrefresh == mode->vrefresh && + hdmi_confs[i].vrefresh == drm_mode_vrefresh(mode) && hdmi_confs[i].interlace == ((mode->flags & DRM_MODE_FLAG_INTERLACE) ? true : false)) diff --git a/drivers/gpu/drm/gma500/psb_drv.c b/drivers/gpu/drm/gma500/psb_drv.c index caba6e0..f2cbcef 100644 --- a/drivers/gpu/drm/gma500/psb_drv.c +++ b/drivers/gpu/drm/gma500/psb_drv.c @@ -505,7 +505,6 @@ static int psb_mode_operation_ioctl(struct drm_device *dev, void *data, mode->vsync_end = umode->vsync_end; mode->vtotal = umode->vtotal; mode->vscan = umode->vscan; - mode->vrefresh = umode->vrefresh; mode->flags = umode->flags; mode->type = umode->type; strncpy(mode->name, umode->name, DRM_DISPLAY_MODE_LEN); diff --git a/drivers/gpu/drm/i2c/ch7006_mode.c b/drivers/gpu/drm/i2c/ch7006_mode.c index c860f24..48f8ede 100644 --- a/drivers/gpu/drm/i2c/ch7006_mode.c +++ b/drivers/gpu/drm/i2c/ch7006_mode.c @@ -121,7 +121,6 @@ struct ch7006_tv_norm_info ch7006_tv_norms[] = { .vscan = 0, \ .flags = DRM_MODE_FLAG_##hsynp##HSYNC | \ DRM_MODE_FLAG_##vsynp##VSYNC, \ - .vrefresh = 0, \ }, \ .enc_hdisp = e_hd, \ .enc_vdisp = e_vd, \ diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c index 7b11edb..d6f99dd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_connector.c +++ b/drivers/gpu/drm/nouveau/nouveau_connector.c @@ -549,7 +549,6 @@ nouveau_connector_native_mode(struct drm_connector *connector) int high_w = 0, high_h = 0, high_v = 0; list_for_each_entry(mode, &nv_connector->base.probed_modes, head) { - mode->vrefresh = drm_mode_vrefresh(mode); if (helper->mode_valid(connector, mode) != MODE_OK || (mode->flags & DRM_MODE_FLAG_INTERLACE)) continue; @@ -570,12 +569,12 @@ nouveau_connector_native_mode(struct drm_connector *connector) continue; if (mode->hdisplay == high_w && mode->vdisplay == high_h && - mode->vrefresh < high_v) + drm_mode_vrefresh(mode) < high_v) continue; high_w = mode->hdisplay; high_h = mode->vdisplay; - high_v = mode->vrefresh; + high_v = drm_mode_vrefresh(mode); largest = mode; } diff --git a/drivers/gpu/drm/nouveau/nv04_crtc.c b/drivers/gpu/drm/nouveau/nv04_crtc.c index 4c31c63..387efb6 100644 --- a/drivers/gpu/drm/nouveau/nv04_crtc.c +++ b/drivers/gpu/drm/nouveau/nv04_crtc.c @@ -315,7 +315,10 @@ nv_crtc_mode_set_vga(struct drm_crtc *crtc, struct drm_display_mode *mode) regp->MiscOutReg = 0x23; /* +hsync +vsync */ } +#if 0 + /* XXX ->clock_index doesn't exist anymore, was always 0 anyway */ regp->MiscOutReg |= (mode->clock_index & 0x03) << 2; +#endif /* * Time Sequencer diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 6b0078f..be63692 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -1893,7 +1893,6 @@ static void vmw_guess_mode_timing(struct drm_display_mode *mode) mode->vtotal = mode->vsync_end + 50; mode->clock = (u32)mode->htotal * (u32)mode->vtotal / 100 * 6; - mode->vrefresh = drm_mode_vrefresh(mode); } @@ -1951,7 +1950,6 @@ int vmw_du_connector_fill_modes(struct drm_connector *connector, mode = drm_mode_duplicate(dev, bmode); if (!mode) return 0; - mode->vrefresh = drm_mode_vrefresh(mode); drm_mode_probed_add(connector, mode); } diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h index d672cf4..d5cf54d 100644 --- a/include/drm/drm_crtc.h +++ b/include/drm/drm_crtc.h @@ -118,7 +118,7 @@ enum drm_mode_status { .hdisplay = (hd), .hsync_start = (hss), .hsync_end = (hse), \ .htotal = (ht), .hskew = (hsk), .vdisplay = (vd), \ .vsync_start = (vss), .vsync_end = (vse), .vtotal = (vt), \ - .vscan = (vs), .flags = (f), .vrefresh = 0 + .vscan = (vs), .flags = (f) #define CRTC_INTERLACE_HALVE_V 0x1 /* halve V values for interlacing */ @@ -151,8 +151,6 @@ struct drm_display_mode { int height_mm; /* Actual mode we give to hw */ - int clock_index; - int synth_clock; int crtc_hdisplay; int crtc_hblank_start; int crtc_hblank_end; @@ -166,16 +164,10 @@ struct drm_display_mode { int crtc_vsync_start; int crtc_vsync_end; int crtc_vtotal; - int crtc_hadjusted; - int crtc_vadjusted; /* Driver private mode info */ - int private_size; int *private; int private_flags; - - int vrefresh; /* in Hz */ - int hsync; /* in kHz */ }; enum drm_connector_status { -- 1.7.7.6

1 0

Re: [Linaro-mm-sig] [RFC] Synchronizing access to buffers shared with dma-buf between drivers/devices
by Erik Gilling 06 Jun '12

06 Jun '12

On Wed, Jun 6, 2012 at 6:33 AM, John Reitan <john.reitan(a)arm.com> wrote: >> But maybe instead of inventing something new, we can just use 'struct >> kthread_work' instead of 'struct kds_callback' plus the two 'void *'s? >> If the user needs some extra args they can embed 'struct >> kthread_work' in their own struct and use container_of() magic in the >> cb. >> >> Plus this is a natural fit if you want to dispatch callbacks instead >> on a kthread_worker, which seems like it would simplify a few things >> when it comes to deadlock avoidance.. ie., not resource deadlock >> avoidance, but dispatching callbacks when some lock is held. > > That sounds like a better approach. > Will make a cleaner API, will look into it. When Tom visited us for android graphics camp in the fall he argued that there were cases where we would want to avoid an extra schedule. Consider the case where the GPU is waiting for a render buffer that the display controller is using. If that render can be kicked off w/o acquiring locks, the display's vsync IRQ handler can call release, which in turn calls the GPU callback, which in turn kicks off the render very quickly w/o having to leave IRQ context. One way around the locking issue with callbacks/async wait is to have async wait return a value to indicate that the resource has been acquired instead of calling the callback. This is the approach I chose in our sync framework. -Erik

1 0

Re: [RFC] Synchronizing access to buffers shared with dma-buf between drivers/devices
by Rob Clark 06 Jun '12

06 Jun '12

Some comments inline.. at this stage mostly superficial issues about how the API works, etc.. not had a chance to dig too much into the implementation yet (although some of my comments about the API would change those anyways). Anyways, thanks for getting the ball rolling on this, and I think I can volunteer linaro to pick up and run w/ this if needed. On Fri, May 25, 2012 at 7:08 PM, Tom Cooksey <tom.cooksey(a)arm.com> wrote: > Hi All, > > I realise it's been a while since this was last discussed, however I'd like > to bring up kernel-side synchronization again. By kernel-side > synchronization, I mean allowing multiple drivers/devices wanting to access > the same buffer to do so without bouncing up to userspace to resolve > dependencies such as "the display controller can't start scanning out a > buffer until the GPU has finished rendering into it". As such, this is > really just an optimization which reduces latency between E.g. The GPU > finishing a rendering job and that buffer being scanned out. I appreciate > this particular example is already solved on desktop graphics cards as the > display controller and 3D core are both controlled by the same driver, so no > "generic" mechanism is needed. However on ARM SoCs, the 3D core (like an ARM > Mali) and display controller tend to be driven by separate drivers, so some > mechanism is needed to allow both drivers to synchronize their access to > buffers. > > There are multiple ways synchronization can be achieved, fences/sync objects > is one common approach, however we're presenting a different approach. > Personally, I quite like fence sync objects, however we believe it requires > a lot of userspace interfaces to be changed to pass around sync object > handles. Our hope is that the kds approach will require less effort to make > use of as no existing userspace interfaces need to be changed. E.g. To use > explicit fences, the struct drm_mode_crtc_page_flip would need a new members > to pass in the handle(s) of sync object(s) which the flip depends on (I.e. > don't flip until these fences fire). The additional benefit of our approach > is that it prevents userspace specifying dependency loops which can cause a > deadlock (see kds.txt for an explanation of what I mean here). > > I have waited until now to bring this up again because I am now able to > share the code I was trying (and failing I think) to explain previously. The > code has now been released under the GPLv2 from ARM Mali's developer portal, > however I've attempted to turn that into a patch to allow it to be discussed > on this list. Please find the patch inline below. > > While KDS defines a very generic mechanism, I am proposing that this code or > at least the concepts be merged with the existing dma_buf code, so a the > struct kds_resource members get moved to struct dma_buf, kds_* functions get > renamed to dma_buf_* functions, etc. So I guess what I'm saying is please > don't review the actual code just yet, only the concepts the code describes, > where kds_resource == dma_duf. > > > Cheers, > > Tom > > > > Author: Tom Cooksey <tom.cooksey(a)arm.com> > Date: Fri May 25 10:45:27 2012 +0100 > > Add new system to allow synchronizing access to resources > > See Documentation/kds.txt for details, however the general > idea is that this kds framework synchronizes multiple drivers > ("clients") wanting to access the same resources, where a > resource is typically a 2D image buffer being shared around > using dma-buf. > > Note: This patch is created by extracting the sources from the > tarball on <http://www.malideveloper.com/open-source-mali-gpus-lin > ux-kernel-device-drivers---dev-releases.php> and putting them in > roughly the right places. > > diff --git a/Documentation/kds.txt b/Documentation/kds.txt fwiw, I think the documentation could be made a bit more generic, but this and code style, etc shouldn't be too hard to fix > new file mode 100644 > index 0000000..a96db21 > --- /dev/null > +++ b/Documentation/kds.txt > @@ -0,0 +1,113 @@ > +# > +# (C) COPYRIGHT 2012 ARM Limited. All rights reserved. > +# > +# This program is free software and is provided to you under the terms of > the GNU General Public License version 2 > +# as published by the Free Software Foundation, and any use by you of this > program is subject to the terms of such GNU licence. > +# > +# A copy of the licence is included with the program, and can also be > obtained from Free Software > +# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA > 02110-1301, USA. > +# > +# > + > + > +============================== > +kds - Kernel Dependency System > +============================== > + > +Introduction > +------------ > +kds provides a mechanism for clients to atomically lock down multiple > abstract resources. > +This can be done either synchronously or asynchronously. > +Abstract resources is used to allow a set of clients to use kds to control > access to any > +resource, an example is structured memory buffers. > + > +kds supports that buffer is locked for exclusive access and sharing of > buffers. > + > +kds can be built as either a integrated feature of the kernel or as a > module. > +It supports being compiled as a module both in-tree and out-of-tree. > + > + > +Concepts > +-------- > +A core concept in kds is abstract resources. > +A kds resource is just an abstraction for some client object, kds doesn't > care what it is. > +Typically EGL will consider UMP buffers as being a resource, thus each UMP > buffer has > +a kds resource for synchronization to the buffer. > + > +kds allows a client to create and destroy the abstract resource objects. > +A new resource object is made available asap (it is just a simple malloc > with some initializations), > +while destroy it requires some external synchronization. > + > +The other core concept in kds is consumer of resources. > +kds is requested to allow a client to consume a set of resources and the > client will be notified when it can consume the resources. > + > +Exclusive access allows only one client to consume a resource. > +Shared access permits multiple consumers to acceess a resource > concurrently. > + > + > +APIs > +---- > +kds provides simple resource allocate and destroy functions. > +Clients use this to instantiate and control the lifetime of the resources > kds manages. > + > +kds provides two ways to wait for resources: > +- Asynchronous wait: the client specifies a function pointer to be called > when wait is over > +- Synchronous wait: Function blocks until access is gained. > + > +The synchronous API has a timeout for the wait. > +The call can early out if a signal is delivered. > + > +After a client is done consuming the resource kds must be notified to > release the resources and let some other client take ownership. > +This is done via resource set release call. > + > +A Windows comparison: > +kds implements WaitForMultipleObjectsEx(..., bWaitAll = TRUE, ...) but also > has an asynchronous version in addition. > +kds resources can be seen as being the same as NT object manager resources. > + > +Internals > +--------- > +kds guarantees atomicity when a set of resources is operated on. > +This is implemented via a global resource lock which is taken by kds when > it updates resource objects. > + > +Internally a resource in kds is a linked list head with some flags. > + > +When a consumer requests access to a set of resources it is queued on each > of the resources. > +The link from the consumer to the resources can be triggered. Once all > links are triggered > +the registered callback is called or the blocking function returns. > +A link is considered triggered if it is the first on the list of consumers > of a resource, > +or if all the links ahead of it is marked as shared and itself is of the > type shared. > + > +When the client is done consuming the consumer object is removed from the > linked lists of > +the resources and a potential new consumer becomes the head of the > resources. > +As we add and remove consumers atomically across all resources we can > guarantee that > +we never introduces a A->B + B->A type of loops/deadlocks. > + > + > +kbase/base implementation > +------------------------- > +A HW job needs access to a set of shared resources. > +EGL tracks this and encodes the set along with the atom in the ringbuffer. > +EGL allocates a (k)base dep object to represent the dependency to the set > of resources and encodes that along with the list of resources. > +This dep object is use to create a dependency from a job chain(atom) to the > resources it needs to run. > +When kbase decodes the atom in the ringbuffer it finds the set of resources > and calls kds to request all the needed resources. > +As EGL needs to know when the kds request is delivered a new base event > object is needed: atom enqueued. This event is only delivered for atoms > which uses kds. > +The callback kbase registers trigger the dependency object described which > would trigger the existing JD system to release the job chain. > +When the atom is done kds resource set release is call to release the > resources. > + > +EGL will typically use exclusive access to the render target, while all > buffers used as input can be marked as shared. > + > + > +Buffer publish/vsync > +-------------------- > +EGL will use a separate ioctl or DRM flip to request the flip. > +If the LCD driver is integrated with kds EGL can do these operations early. > +The LCD driver must then implement the ioctl or DRM flip to be asynchronous > with kds async call. > +The LCD driver binds a kds resource to each virtual buffer (2 buffers in > case of double-buffering). > +EGL will make a dependency to the target kds resource in the kbase atom. > +After EGL receives a atom enqueued event it can ask the LCD driver to pan > to the target kds resource. > +When the atom is completed it'll release the resource and the LCD driver > will get its callback. > +In the callback it'll load the target buffer into the DMA unit of the LCD > hardware. > +The LCD driver will be the consumer of both buffers for a short period. > +The LCD driver will call kds resource set release on the previous on-screen > buffer when the next vsync/dma read end is handled. > + > + > diff --git a/drivers/misc/kds.c b/drivers/misc/kds.c > new file mode 100644 > index 0000000..8d7d55e > --- /dev/null > +++ b/drivers/misc/kds.c > @@ -0,0 +1,461 @@ > +/* > + * > + * (C) COPYRIGHT 2012 ARM Limited. All rights reserved. > + * > + * This program is free software and is provided to you under the terms of > the GNU General Public License version 2 > + * as published by the Free Software Foundation, and any use by you of this > program is subject to the terms of such GNU licence. > + * > + * A copy of the licence is included with the program, and can also be > obtained from Free Software > + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA > 02110-1301, USA. > + * > + */ > + > + > + > +#include <linux/slab.h> > +#include <linux/list.h> > +#include <linux/mutex.h> > +#include <linux/wait.h> > +#include <linux/sched.h> > +#include <linux/err.h> > +#include <linux/module.h> > +#include <linux/workqueue.h> > +#include <linux/kds.h> > + > + > +#define KDS_LINK_TRIGGERED (1u << 0) > +#define KDS_LINK_EXCLUSIVE (1u << 1) > + > +#define KDS_IGNORED NULL > +#define KDS_INVALID (void*)-2 > +#define KDS_RESOURCE (void*)-1 > + > +struct kds_resource_set > +{ > + unsigned long num_resources; > + unsigned long pending; > + unsigned long locked_resources; > + struct kds_callback * cb; > + void * callback_parameter; > + void * callback_extra_parameter; > + struct list_head callback_link; > + struct work_struct callback_work; > + struct kds_link resources[0]; > +}; > + > +static DEFINE_MUTEX(kds_lock); > + > +int kds_callback_init(struct kds_callback * cb, int direct, kds_callback_fn > user_cb) > +{ > + int ret = 0; > + > + cb->direct = direct; > + cb->user_cb = user_cb; > + > + if (!direct) > + { > + cb->wq = alloc_workqueue("kds", WQ_UNBOUND | WQ_HIGHPRI, > WQ_UNBOUND_MAX_ACTIVE); > + if (!cb->wq) > + ret = -ENOMEM; > + } > + else > + { > + cb->wq = NULL; > + } > + > + return ret; > +} > +EXPORT_SYMBOL(kds_callback_init); > + > +void kds_callback_term(struct kds_callback * cb) > +{ > + if (!cb->direct) > + { > + BUG_ON(!cb->wq); > + destroy_workqueue(cb->wq); > + } > + else > + { > + BUG_ON(cb->wq); > + } > +} > + > +EXPORT_SYMBOL(kds_callback_term); > + > +static void kds_do_user_callback(struct kds_resource_set * rset) > +{ > + rset->cb->user_cb(rset->callback_parameter, > rset->callback_extra_parameter); > +} > + > +static void kds_queued_callback(struct work_struct * work) > +{ > + struct kds_resource_set * rset; > + rset = container_of( work, struct kds_resource_set, callback_work); > + > + kds_do_user_callback(rset); > +} > + > +static void kds_callback_perform(struct kds_resource_set * rset) > +{ > + if (rset->cb->direct) > + kds_do_user_callback(rset); > + else > + { > + int result; > + result = queue_work(rset->cb->wq, &rset->callback_work); > + /* if we got a 0 return it means we've triggered the same > rset twice! */ > + BUG_ON(!result); > + } > +} > + > +void kds_resource_init(struct kds_resource * res) > +{ > + BUG_ON(!res); > + INIT_LIST_HEAD(&res->waiters.link); > + res->waiters.parent = KDS_RESOURCE; > +} > +EXPORT_SYMBOL(kds_resource_init); > + > +void kds_resource_term(struct kds_resource * res) > +{ > + BUG_ON(!res); > + BUG_ON(!list_empty(&res->waiters.link)); > + res->waiters.parent = KDS_INVALID; > +} > +EXPORT_SYMBOL(kds_resource_term); > + > +int kds_async_waitall( > + struct kds_resource_set ** pprset, > + unsigned long flags, > + struct kds_callback * cb, > + void * callback_parameter, > + void * callback_extra_parameter, > + int number_resources, > + unsigned long * exclusive_access_bitmap, > + struct kds_resource ** resource_list) > +{ > + struct kds_resource_set * rset = NULL; > + int i; > + int triggered; > + int err = -EFAULT; > + > + BUG_ON(!pprset); > + BUG_ON(!resource_list); > + BUG_ON(!cb); > + > + mutex_lock(&kds_lock); > + > + if ((flags & KDS_FLAG_LOCKED_ACTION) == KDS_FLAG_LOCKED_FAIL) > + { > + for (i = 0; i < number_resources; i++) > + { > + if (resource_list[i]->lock_count) > + { > + err = -EBUSY; > + goto errout; > + } > + } > + } > + > + rset = kmalloc(sizeof(*rset) + number_resources * sizeof(struct > kds_link), GFP_KERNEL); > + if (!rset) > + { > + err = -ENOMEM; > + goto errout; > + } > + > + rset->num_resources = number_resources; > + rset->pending = number_resources; > + rset->locked_resources = 0; > + rset->cb = cb; > + rset->callback_parameter = callback_parameter; > + rset->callback_extra_parameter = callback_extra_parameter; > + INIT_LIST_HEAD(&rset->callback_link); > + INIT_WORK(&rset->callback_work, kds_queued_callback); > + > + for (i = 0; i < number_resources; i++) > + { > + unsigned long link_state = 0; > + > + INIT_LIST_HEAD(&rset->resources[i].link); > + rset->resources[i].parent = rset; > + > + if (test_bit(i, exclusive_access_bitmap)) > + { > + link_state |= KDS_LINK_EXCLUSIVE; > + } > + > + /* no-one else waiting? */ > + if (list_empty(&resource_list[i]->waiters.link)) > + { > + link_state |= KDS_LINK_TRIGGERED; > + rset->pending--; > + } > + /* Adding a non-exclusive and the current tail is a > triggered non-exclusive? */ > + else if (((link_state & KDS_LINK_EXCLUSIVE) == 0) && > + (((list_entry(resource_list[i]->waiters.link.prev, > struct kds_link, link)->state & (KDS_LINK_EXCLUSIVE | KDS_LINK_TRIGGERED)) > == KDS_LINK_TRIGGERED))) > + { > + link_state |= KDS_LINK_TRIGGERED; > + rset->pending--; > + } > + /* locked & ignore locked? */ > + else if ((resource_list[i]->lock_count) && ((flags & > KDS_FLAG_LOCKED_ACTION) == KDS_FLAG_LOCKED_IGNORE) ) > + { > + link_state |= KDS_LINK_TRIGGERED; > + rset->pending--; > + rset->resources[i].parent = KDS_IGNORED; /* to > disable decrementing the pending count when we get the ignored resource */ > + } > + rset->resources[i].state = link_state; > + list_add_tail(&rset->resources[i].link, > &resource_list[i]->waiters.link); > + } > + > + triggered = (rset->pending == 0); > + > + mutex_unlock(&kds_lock); > + > + /* set the pointer before the callback is called so it sees it */ > + *pprset = rset; > + > + if (triggered) > + { > + /* all resources obtained, trigger callback */ > + kds_callback_perform(rset); > + } > + > + return 0; > + > +errout: > + mutex_unlock(&kds_lock); > + return err; > +} > +EXPORT_SYMBOL(kds_async_waitall); > + > +static void wake_up_sync_call(void * callback_parameter, void * > callback_extra_parameter) > +{ > + wait_queue_head_t * wait = (wait_queue_head_t*)callback_parameter; > + wake_up(wait); > +} > + > +static struct kds_callback sync_cb = > +{ > + wake_up_sync_call, > + 1, > + NULL, > +}; > + > +struct kds_resource_set * kds_waitall( > + int number_resources, > + unsigned long * exclusive_access_bitmap, > + struct kds_resource ** resource_list, > + unsigned long jiffies_timeout) > +{ > + struct kds_resource_set * rset; > + int i; > + int triggered = 0; > + DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wake); > + > + rset = kmalloc(sizeof(*rset) + number_resources * sizeof(struct > kds_link), GFP_KERNEL); > + if (!rset) > + return rset; > + > + rset->num_resources = number_resources; > + rset->pending = number_resources; > + rset->locked_resources = 1; > + INIT_LIST_HEAD(&rset->callback_link); > + INIT_WORK(&rset->callback_work, kds_queued_callback); > + > + mutex_lock(&kds_lock); > + > + for (i = 0; i < number_resources; i++) > + { > + unsigned long link_state = 0; > + > + if (likely(resource_list[i]->lock_count < ULONG_MAX)) > + resource_list[i]->lock_count++; > + else > + break; > + > + if (test_bit(i, exclusive_access_bitmap)) > + { > + link_state |= KDS_LINK_EXCLUSIVE; > + } > + > + if (list_empty(&resource_list[i]->waiters.link)) > + { > + link_state |= KDS_LINK_TRIGGERED; > + rset->pending--; > + } > + /* Adding a non-exclusive and the current tail is a > triggered non-exclusive? */ > + else if (((link_state & KDS_LINK_EXCLUSIVE) == 0) && > + (((list_entry(resource_list[i]->waiters.link.prev, > struct kds_link, link)->state & (KDS_LINK_EXCLUSIVE | KDS_LINK_TRIGGERED)) > == KDS_LINK_TRIGGERED))) > + { > + link_state |= KDS_LINK_TRIGGERED; > + rset->pending--; > + } > + > + INIT_LIST_HEAD(&rset->resources[i].link); > + rset->resources[i].parent = rset; > + rset->resources[i].state = link_state; > + list_add_tail(&rset->resources[i].link, > &resource_list[i]->waiters.link); > + } > + > + if (i < number_resources) > + { > + /* an overflow was detected, roll back */ > + while (i--) > + { > + list_del(&rset->resources[i].link); > + resource_list[i]->lock_count--; > + } > + mutex_unlock(&kds_lock); > + kfree(rset); > + return ERR_PTR(-EFAULT); > + } > + > + if (rset->pending == 0) > + triggered = 1; > + else > + { > + rset->cb = &sync_cb; > + rset->callback_parameter = &wake; > + rset->callback_extra_parameter = NULL; > + } > + > + mutex_unlock(&kds_lock); > + > + if (!triggered) > + { > + long wait_res; > + if ( KDS_WAIT_BLOCKING == jiffies_timeout ) > + { > + wait_res = wait_event_interruptible(wake, > rset->pending == 0); > + } > + else > + { > + wait_res = wait_event_interruptible_timeout(wake, > rset->pending == 0, jiffies_timeout); > + } > + if ((wait_res == -ERESTARTSYS) || (wait_res == 0)) > + { > + /* use \a kds_resource_set_release to roll back */ > + kds_resource_set_release(&rset); > + return ERR_PTR(wait_res); > + } > + } > + return rset; > +} > +EXPORT_SYMBOL(kds_waitall); > + > +void kds_resource_set_release(struct kds_resource_set ** pprset) > +{ > + struct list_head triggered = LIST_HEAD_INIT(triggered); > + struct kds_resource_set * rset; > + struct kds_resource_set * it; > + int i; > + > + BUG_ON(!pprset); > + > + mutex_lock(&kds_lock); > + > + rset = *pprset; > + if (!rset) > + { > + /* caught a race between a cancelation > + * and a completion, nothing to do */ > + mutex_unlock(&kds_lock); > + return; > + } > + > + /* clear user pointer so we'll be the only > + * thread handling the release */ > + *pprset = NULL; > + > + for (i = 0; i < rset->num_resources; i++) > + { > + struct kds_resource * resource; > + struct kds_link * it = NULL; > + > + /* fetch the previous entry on the linked list */ > + it = list_entry(rset->resources[i].link.prev, struct > kds_link, link); > + /* unlink ourself */ > + list_del(&rset->resources[i].link); > + > + /* any waiters? */ > + if (list_empty(&it->link)) > + continue; > + > + /* were we the head of the list? (head if prev is a > resource) */ > + if (it->parent != KDS_RESOURCE) > + continue; > + > + /* we were the head, find the kds_resource */ > + resource = container_of(it, struct kds_resource, waiters); > + > + if (rset->locked_resources) > + { > + resource->lock_count--; > + } > + > + /* we know there is someone waiting from the any-waiters > test above */ > + > + /* find the head of the waiting list */ > + it = list_first_entry(&resource->waiters.link, struct > kds_link, link); > + > + /* new exclusive owner? */ > + if (it->state & KDS_LINK_EXCLUSIVE) > + { > + /* link now triggered */ > + it->state |= KDS_LINK_TRIGGERED; > + /* a parent to update? */ > + if (it->parent != KDS_IGNORED) > + { > + if (0 == --it->parent->pending) > + { > + /* new owner now triggered, track > for callback later */ > + list_add(&it->parent->callback_link, > &triggered); > + } > + } > + } > + /* exclusive releasing ? */ > + else if (rset->resources[i].state & KDS_LINK_EXCLUSIVE) > + { > + /* trigger non-exclusive until end-of-list or first > exclusive */ > + list_for_each_entry(it, &resource->waiters.link, > link) > + { > + /* exclusive found, stop triggering */ > + if (it->state & KDS_LINK_EXCLUSIVE) > + break; > + > + it->state |= KDS_LINK_TRIGGERED; > + /* a parent to update? */ > + if (it->parent != KDS_IGNORED) > + { > + if (0 == --it->parent->pending) > + { > + /* new owner now triggered, > track for callback later */ > + > list_add(&it->parent->callback_link, &triggered); > + } > + } > + } > + } > + > + } > + > + mutex_unlock(&kds_lock); > + > + while (!list_empty(&triggered)) > + { > + it = list_first_entry(&triggered, struct kds_resource_set, > callback_link); > + list_del(&it->callback_link); > + kds_callback_perform(it); > + } > + > + cancel_work_sync(&rset->callback_work); > + > + /* free the resource set */ > + kfree(rset); > +} > +EXPORT_SYMBOL(kds_resource_set_release); > + > +MODULE_LICENSE("GPL"); > +MODULE_AUTHOR("ARM Ltd."); > +MODULE_VERSION("1.0"); > diff --git a/include/linux/kds.h b/include/linux/kds.h > new file mode 100644 > index 0000000..65e5706 > --- /dev/null > +++ b/include/linux/kds.h > @@ -0,0 +1,154 @@ > +/* > + * > + * (C) COPYRIGHT 2012 ARM Limited. All rights reserved. > + * > + * This program is free software and is provided to you under the terms of > the GNU General Public License version 2 > + * as published by the Free Software Foundation, and any use by you of this > program is subject to the terms of such GNU licence. > + * > + * A copy of the licence is included with the program, and can also be > obtained from Free Software > + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA > 02110-1301, USA. > + * > + */ > + > + > + > +#ifndef _KDS_H_ > +#define _KDS_H_ > + > +#include <linux/list.h> > +#include <linux/workqueue.h> > + > +#define KDS_WAIT_BLOCKING (ULONG_MAX) > + > +/* what to do when waitall must wait for a synchronous locked resource: */ > +#define KDS_FLAG_LOCKED_FAIL (0u << 0) /* fail waitall */ > +#define KDS_FLAG_LOCKED_IGNORE (1u << 0) /* don't wait, but block other > that waits */ > +#define KDS_FLAG_LOCKED_WAIT (2u << 0) /* wait (normal */ > +#define KDS_FLAG_LOCKED_ACTION (3u << 0) /* mask to extract the action to > do on locked resources */ > + > +struct kds_resource_set; > + > +typedef void (*kds_callback_fn) (void * callback_parameter, void * > callback_extra_parameter); > + > +struct kds_callback > +{ > + kds_callback_fn user_cb; /* real cb */ > + int direct; /* do direct or queued call? */ > + struct workqueue_struct * wq; > +}; > + > +struct kds_link > +{ > + struct kds_resource_set * parent; > + struct list_head link; > + unsigned long state; > +}; > + > +struct kds_resource > +{ > + struct kds_link waiters; > + unsigned long lock_count; > +}; > + > +/* callback API */ > + > +/* Initialize a callback object. > + * > + * Typically created per context or per hw resource. > + * > + * Callbacks can be performed directly if no nested locking can > + * happen in the client. > + * > + * Nested locking can occur when a lock is held during the > kds_async_waitall or > + * kds_resource_set_release call. If the callback needs to take the same > lock > + * nested locking will happen. > + * > + * If nested locking could happen non-direct callbacks can be requested. > + * Callbacks will then be called asynchronous to the triggering call. > + */ > +int kds_callback_init(struct kds_callback * cb, int direct, kds_callback_fn > user_cb); > + > +/* Terminate the use of a callback object. > + * > + * If the callback object was set up as non-direct > + * any pending callbacks will be flushed first. > + * Note that to avoid a deadlock the lock callbacks needs > + * can't be held when a callback object is terminated. > + */ > +void kds_callback_term(struct kds_callback * cb); hmm, not hugely a fan of this.. having callbacks that might need to aquire locks be potentially called synchronously in special cases. It seems like it can get simpler if pending callbacks could hold a reference to the underlying object until the callback completes. Although not quite sure offhand how that can work without coupling kds to dmabuf or GEM.. > + > +/* resource object API */ > + > +/* initialize a resource handle for a shared resource */ > +void kds_resource_init(struct kds_resource * resource); > + > +/* > + * Will assert if the resource is being used or waited on. > + * The caller should NOT try to terminate a resource that could still have > clients. > + * After the function returns the resource is no longer known by kds. > + */ > +void kds_resource_term(struct kds_resource * resource); > + > +/* Asynchronous wait for a set of resources. > + * Callback will be called when all resources are available. > + * If all the resources was available the callback will be called before > kds_async_waitall returns. > + * So one must not hold any locks the callback code-flow can take when > calling kds_async_waitall. > + * Caller considered to own/use the resources until \a kds_rset_release is > called. > + * flags is one or more of the KDS_FLAG_* set. > + * exclusive_access_bitmap is a bitmap where a high bit means exclusive > access while a low bit means shared access. > + * Use the Linux __set_bit API, where the index of the buffer to control is > used as the bit index. > + * > + * Standard Linux error return value. > + */ > +int kds_async_waitall( > + struct kds_resource_set ** pprset, > + unsigned long flags, > + struct kds_callback * cb, > + void * callback_parameter, > + void * callback_extra_parameter, > + int number_resources, > + unsigned long * exclusive_access_bitmap, hmm, is there advantage of passing the requested resources this way, vs. just having two arrays (one for exclusive access, one for shared access)? > + struct kds_resource ** resource_list); callback_parameter + callback_extra_parameter seems a bit, well, odd. I'm guessing that is some implementation detail about mali driver peaking through? But maybe instead of inventing something new, we can just use 'struct kthread_work' instead of 'struct kds_callback' plus the two 'void *'s? If the user needs some extra args they can embed 'struct kthread_work' in their own struct and use container_of() magic in the cb. Plus this is a natural fit if you want to dispatch callbacks instead on a kthread_worker, which seems like it would simplify a few things when it comes to deadlock avoidance.. ie., not resource deadlock avoidance, but dispatching callbacks when some lock is held. /me wonders what sort of fun would otherwise happen if the cb ever indirectly called kds_resource_set_release()? > +/* Synchronous wait for a set of resources. > + * Function will return when one of these have happened: > + * - all resources have been obtained > + * - timeout lapsed while waiting > + * - a signal was received while waiting > + * > + * Caller considered to own/use the resources when the function returns. > + * Caller must release the resources using \a kds_rset_release. > + * > + * Calling this function while holding already locked resources or other > locking primitives is dangerous. > + * One must if this is needed decide on a lock order of the resources > and/or the other locking primitives > + * and always take the resources/locking primitives in the specific order. > + * > + * Use the ERR_PTR framework to decode the return value. > + * NULL = time out > + * If IS_ERR then PTR_ERR gives: > + * ERESTARTSYS = signal received, retry call after signal > + * all other values = internal error, lock failed > + * Other values = successful wait, now the owner, must call > kds_resource_set_release > + */ > +struct kds_resource_set * kds_waitall( > + int number_resources, > + unsigned long * exclusive_access_bitmap, > + struct kds_resource ** resource_list, > + unsigned long jifies_timeout); > + > +/* Release resources after use. > + * Caller must handle that other async callbacks will trigger, > + * so must avoid holding any locks a callback will take. > + * > + * The function takes a pointer to your poiner to handle a race > + * between a cancelation and a completion. > + * > + * If the caller can't guarantee that a race can't occur then > + * the passed in pointer must be the same in both call paths > + * to allow kds to manage the potential race. > + */ > +void kds_resource_set_release(struct kds_resource_set ** pprset); maybe using a worker as mentioned above for dealing w/ async cb's simplify's things? Maybe I'm a bit paranoid about locking around all of this, but basically the synchronization framework ends up having to deal w/ all the potential deadlock and recursive lock issues that we tried to hide from w/ dmabuf ;-) BR, -R > + > +#endif /* _KDS_H_ */ > + > > > > _______________________________________________ > dri-devel mailing list > dri-devel(a)lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/dri-devel

2 1

[PATCH] radeon: fall back to 1D tiling only with broken kernels
by alexdeucher＠gmail.com 06 Jun '12

06 Jun '12

From: Alex Deucher <alexander.deucher(a)amd.com> Certain cards report the the wrong bank setup which causes surface init to fail in the ddx and leads to no accel. If we hit an invalid tiling parameter, just set a default value and disable 2D tiling. Should fix: https://bugs.freedesktop.org/show_bug.cgi?id=43448 Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> --- radeon/radeon_surface.c | 28 +++++++++++++++++++++------- 1 files changed, 21 insertions(+), 7 deletions(-) diff --git a/radeon/radeon_surface.c b/radeon/radeon_surface.c index a7b488b..adf209d 100644 --- a/radeon/radeon_surface.c +++ b/radeon/radeon_surface.c @@ -206,7 +206,9 @@ static int r6_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.num_pipes = 8; break; default: - return -EINVAL; + surf_man->hw_info.num_pipes = 8; + surf_man->hw_info.allow_2d = 0; + break; } switch ((tiling_config & 0x30) >> 4) { @@ -217,7 +219,9 @@ static int r6_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.num_banks = 8; break; default: - return -EINVAL; + surf_man->hw_info.num_banks = 8; + surf_man->hw_info.allow_2d = 0; + break; } switch ((tiling_config & 0xc0) >> 6) { @@ -228,7 +232,9 @@ static int r6_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.group_bytes = 512; break; default: - return -EINVAL; + surf_man->hw_info.group_bytes = 256; + surf_man->hw_info.allow_2d = 0; + break; } return 0; } @@ -458,7 +464,9 @@ static int eg_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.num_pipes = 8; break; default: - return -EINVAL; + surf_man->hw_info.num_pipes = 8; + surf_man->hw_info.allow_2d = 0; + break; } switch ((tiling_config & 0xf0) >> 4) { @@ -472,7 +480,9 @@ static int eg_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.num_banks = 16; break; default: - return -EINVAL; + surf_man->hw_info.num_banks = 8; + surf_man->hw_info.allow_2d = 0; + break; } switch ((tiling_config & 0xf00) >> 8) { @@ -483,7 +493,9 @@ static int eg_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.group_bytes = 512; break; default: - return -EINVAL; + surf_man->hw_info.group_bytes = 256; + surf_man->hw_info.allow_2d = 0; + break; } switch ((tiling_config & 0xf000) >> 12) { @@ -497,7 +509,9 @@ static int eg_init_hw_info(struct radeon_surface_manager *surf_man) surf_man->hw_info.row_size = 4096; break; default: - return -EINVAL; + surf_man->hw_info.row_size = 4096; + surf_man->hw_info.allow_2d = 0; + break; } return 0; } -- 1.7.7.5

1 0

← Newer
1
...
63
64
65
66
67
68
69
70
Older →

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

dri-devel June 2012