dri-devel December 2012

dri-devel@lists.freedesktop.org

121 participants
687 discussions

[Bug 37485] New: Noise in some tracks in SuperTuxKart (HyperZ related)
by bugzilla-daemon＠freedesktop.org 04 Dec '12

04 Dec '12

https://bugs.freedesktop.org/show_bug.cgi?id=37485 Summary: Noise in some tracks in SuperTuxKart (HyperZ related) Product: Mesa Version: git Platform: x86 (IA32) OS/Version: Linux (All) Status: NEW Severity: normal Priority: medium Component: Drivers/DRI/r300 AssignedTo: dri-devel(a)lists.freedesktop.org ReportedBy: gsr.bugs(a)infernal-iceberg.com Some tracks of STK show noise in two forms. One is thin lines, in yellow over white surface, for example, like if you were modelling the track and wanted to see the polygons edges. And the second is small squares or rectangles where karts' shadows appear or also near polygon edges, where walls and grounds meet, for example. It's rather hard to capture on a picture (slow computer, karts sliding/shaking a bit even if stopped, so the noise flickers constantly) but the square one is like some HyperZ bugs, you can guess the colour comes from other colours in the same track. When RADEON_HYPERZ is not set, things render as they should. This is in a RV360 card, not R5xx like the other reports. -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.

1 3

[PATCH] drm/radeon: fix eDP clk and lane setup for scaled modes
by alexdeucher＠gmail.com 04 Dec '12

04 Dec '12

From: Alex Deucher <alexander.deucher(a)amd.com> Need to use the adjusted mode since we are sending native timing and using the scaler for non-native modes. Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> cc: stable(a)vger.kernel.org --- drivers/gpu/drm/radeon/atombios_encoders.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/drivers/gpu/drm/radeon/atombios_encoders.c b/drivers/gpu/drm/radeon/atombios_encoders.c index 010bae1..4552d4a 100644 --- a/drivers/gpu/drm/radeon/atombios_encoders.c +++ b/drivers/gpu/drm/radeon/atombios_encoders.c @@ -340,7 +340,7 @@ static bool radeon_atom_mode_fixup(struct drm_encoder *encoder, ((radeon_encoder->active_device & (ATOM_DEVICE_DFP_SUPPORT | ATOM_DEVICE_LCD_SUPPORT)) || (radeon_encoder_get_dp_bridge_encoder_id(encoder) != ENCODER_OBJECT_ID_NONE))) { struct drm_connector *connector = radeon_get_connector_for_encoder(encoder); - radeon_dp_set_link_config(connector, mode); + radeon_dp_set_link_config(connector, adjusted_mode); } return true; -- 1.7.7.5

2 1

[Bug 51301] [-next-20121129] memory leak/page allocator fragmentation?
by bugzilla-daemon＠bugzilla.kernel.org 04 Dec '12

04 Dec '12

1 0

[Bug 51301] [-next-20121129] memory leak/page allocator fragmentation?
by bugzilla-daemon＠bugzilla.kernel.org 04 Dec '12

04 Dec '12

https://bugzilla.kernel.org/show_bug.cgi?id=51301 Marcelo Tosatti <mtosatti(a)redhat.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mtosatti(a)redhat.com --- Comment #5 from Marcelo Tosatti <mtosatti(a)redhat.com> 2012-12-04 21:42:00 --- Dec 4 09:38:25 thor kernel: [322576.251464] firefox: page allocation failure: order:4, mode:0x80c0d0 This is an order-4 allocation failure. This is due to kernel failure to allocate physically contiguous region of memory. Please reassign to "Memory Management" component. -- Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching someone on the CC list of the bug.

1 0

[Bug 51301] [-next-20121129] memory leak/page allocator fragmentation?
by bugzilla-daemon＠bugzilla.kernel.org 04 Dec '12

04 Dec '12

https://bugzilla.kernel.org/show_bug.cgi?id=51301 --- Comment #4 from Peter Hurley <peter(a)hurleysoftware.com> 2012-12-04 16:48:22 --- Not sure why I picked kvm/nouveau for this OOM ;) Could be any number of apps over a 3.5 period... -- Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching someone on the CC list of the bug.

1 0

[RFC 0/4] drm/exynos: add ipp subsystem and each ipp drivers
by Eunchul Kim 04 Dec '12

04 Dec '12

Hi All. I am responsible for a display part from Samsung Electronics Telecommunication Division. and I am going to add post-processing features in exynos drm. If you have some opinions of this patch, please give some comments about my patch. This patch is post-processing(IPP) support for exynos drm driver. IPP is stands for Image Post Processing and supports image scaler/rotator and input/output DMA operations using IPP drivers(FIMC, Rotator, GSC, SC, so on.) IPP is integration device driver of same attibute hardware. Exynos4xxxx SoC support FIMC, Rotator for post-processing. and Exynos5xxxx SoC support GSC, Rotator, SC for post-processing. SC driver not tested yet. so, I didn't add SC feature in this patch. and IPP subsystem works on Exynos SoC version independently. IPP drivers supports Memory to Memory operations with various converting. and some drivers supports Writeback and Display output operations using local path. User can make converted image using this driver. and also supports streaming concept for multimedia data processing. And supports various operations 1. Scale operation generates various sizes of image. 2. CSC(Color Space Conversion) operation supports format converting. 3. Crop operation supports cutting the image. 4. Rotate operation supports to 90, 180, 270 degree. 5. Flip operation supports to vertical, horizontal, and both. 6. Writeback operation genertates cloned image from display controller(FIMD). 7. Display output operation support various format display to display controller(FIMD). 8. Input DMA reads image data from the memory. 9. Output DMA writes image data to memory. 10. Supports image effect functions. Descriptions) User should make property informations and set this property to registers. and IPP subsystem manages property id using command node and make queue list using memory node. IPP subsystem supports register function of IPP drivers. IPP driver manages input/output DMA with various operations. and some driver supports optional operations(writeback, output). IPP driver needs various informations, so User set property information to IPP driver. and also IPP driver needs memory base address of image for various operations. User doesn't know its address, so fills the gem handle of that memory than address of image base. and than IPP driver start operation. Ioctls) We adds four ioctls and one event for IPP subsystem. - ioctls DRM_EXYNOS_IPP_GET_PROPERTY : get ipp driver capabilitis and id. DRM_EXYNOS_IPP_SET_PROPERTY : set format, position, rotation, flip about source/destination. DRM_EXYNOS_IPP_QUEUE_BUF : enqueue/dequeue buffer and make event list. DRM_EXYNOS_IPP_CMD_CTRL : play/stop/pause/resume control. - event DRM_EXYNOS_IPP_EVENT : event to give notification completion of buffer DMA with buffer list Basic control flow and Sample pseudo application) 1. Basic control flow is same as below Open -> Get properties -> User choose IPP driver and set property information -> Set Property -> Create gem handle -> Queue buffer(Enqueue) of source/destination -> Command control(Play) -> Event occured to User -> User handle converted image -> (Queue buffer(Enqueue) of source/destination -> Event occured to User)*N -> Queue buffer(Dequeue) of source/destination -> Command control(Stop) -> Free gem handle -> Close 2. Sample pseudo application #include "exynos_drm.h" static int exynos_drm_ipp_set_property(int fd ...) { struct drm_exynos_pos crop_pos = {0, 0, hsize, vsize}; struct drm_exynos_pos scale_pos = {0, 0, hsize, vsize}; struct drm_exynos_sz src_sz = {hsize, vsize}; struct drm_exynos_sz dst_sz = {hsize, vsize}; int ret = 0; memset(property, 0x00, sizeof(struct drm_exynos_ipp_property)); property->cmd = cmd; property->config[EXYNOS_DRM_OPS_SRC].ops_id = EXYNOS_DRM_OPS_SRC; property->config[EXYNOS_DRM_OPS_SRC].flip = EXYNOS_DRM_FLIP_NONE; property->config[EXYNOS_DRM_OPS_SRC].degree = EXYNOS_DRM_DEGREE_0; property->config[EXYNOS_DRM_OPS_SRC].fmt = DRM_FORMAT_XRGB8888; property->config[EXYNOS_DRM_OPS_SRC].pos = crop_pos; property->config[EXYNOS_DRM_OPS_SRC].sz = src_sz; property->config[EXYNOS_DRM_OPS_DST].ops_id = EXYNOS_DRM_OPS_DST; property->config[EXYNOS_DRM_OPS_DST].flip = EXYNOS_DRM_FLIP_NONE; property->config[EXYNOS_DRM_OPS_DST].degree = degree; property->config[EXYNOS_DRM_OPS_DST].fmt = DRM_FORMAT_XRGB8888; property->config[EXYNOS_DRM_OPS_DST].pos = scale_pos; property->config[EXYNOS_DRM_OPS_DST].sz = dst_sz; ret = ioctl(fd, DRM_IOCTL_EXYNOS_IPP_SET_PROPERTY, property); if (ret) fprintf(stderr, "failed to DRM_IOCTL_EXYNOS_IPP_SET_PROPERTY : %s\n", strerror(errno)); printf("DRM_IOCTL_EXYNOS_IPP_SET_PROPERTY : prop_id[%d]\n", property->prop_id); return ret; } static int exynos_drm_ipp_queue_buf(int fd, ..., unsigned int gem_handle) { int ret = 0; memset(qbuf, 0x00, sizeof(struct drm_exynos_ipp_queue_buf)); qbuf->ops_id = ops_id; qbuf->buf = buf; qbuf->user_data = 0; qbuf->prop_id = prop_id; qbuf->buf_id = buf_id; qbuf->handle[EXYNOS_DRM_PLANAR_Y] = gem_handle; qbuf->handle[EXYNOS_DRM_PLANAR_CB] = 0; qbuf->handle[EXYNOS_DRM_PLANAR_CR] = 0; ret = ioctl(fd, DRM_IOCTL_EXYNOS_IPP_QUEUE_BUF, qbuf); if (ret) fprintf(stderr, "failed to DRM_IOCTL_EXYNOS_IPP_QUEUE_BUF[id:%d][buf:%d] : %s\n", ops_id, buf, strerror(errno)); return ret; } static int exynos_drm_ipp_cmd_ctrl(int fd, ...) { int ret = 0; memset(cmd_ctrl, 0x00, sizeof(struct drm_exynos_ipp_cmd_ctrl)); cmd_ctrl->prop_id = prop_id; cmd_ctrl->ctrl = ctrl; ret = ioctl(fd, DRM_IOCTL_EXYNOS_IPP_CMD_CTRL, cmd_ctrl); if (ret) fprintf(stderr, "failed to DRM_IOCTL_EXYNOS_IPP_CMD_CTRL[prop_id:%d][ctrl:%d] : %s\n", prop_id, ctrl, strerror(errno)); return ret; } int ipp_event_handler(...) { char buffer[1024]; int len, i; struct drm_event *e; struct drm_exynos_ipp_event *ipp_event; char filename[100]; int ret = 0; int src_buf_id, dst_buf_id; static bmp_idx = 0; len = read(fd, buffer, sizeof buffer); if (len == 0) return 0; if (len < sizeof *e) return -1; i = 0; while (i < len) { e = (struct drm_event *) &buffer[i]; switch (e->type) { case DRM_EXYNOS_IPP_EVENT: ipp_event = (struct drm_exynos_ipp_event *) e; src_buf_id = ipp_event->buf_id[EXYNOS_DRM_OPS_SRC]; dst_buf_id = ipp_event->buf_id[EXYNOS_DRM_OPS_DST]; /* For source buffer queue to IPP */ ret = exynos_drm_ipp_queue_buf(fd, &src_qbuf[src_buf_id], EXYNOS_DRM_OPS_SRC, IPP_BUF_ENQUEUE, property->prop_id, src_buf_id, src_gem[src_buf_id].handle); if (ret) { fprintf(stderr, "failed to ipp buf src queue\n"); goto err_ipp_ctrl_close; } /* For destination buffer queue to IPP */ ret = exynos_drm_ipp_queue_buf(fd, &dst_qbuf[dst_buf_id], EXYNOS_DRM_OPS_DST, IPP_BUF_ENQUEUE, property->prop_id, dst_buf_id, dst_gem[dst_buf_id].handle); if (ret) { fprintf(stderr, "failed to ipp buf dst queue\n"); goto err_ipp_ctrl_close; } break; default: break; } i += e->length; } err_ipp_ctrl_close: return ret; } void ipp_main(...) { struct drm_exynos_ipp_property property; struct drm_exynos_ipp_cmd_ctrl cmd_ctrl; struct drm_exynos_sz def_sz = {720, 1280}; struct drm_exynos_ipp_queue_buf qbuf1[MAX_BUF], qbuf2[MAX_BUF]; unsigned int width=720, height=1280, stride; int ret, i, j, x; struct drm_exynos_gem_create gem1[MAX_BUF], gem2[MAX_BUF]; struct drm_exynos_gem_mmap mmap1[MAX_BUF], mmap2[MAX_BUF]; void *usr_addr1[MAX_BUF], *usr_addr2[MAX_BUF]; struct timeval begin, end; struct drm_gem_close args; char filename[100]; /* For property */ ret = exynos_drm_ipp_set_property(fd, &property, &def_sz, IPP_CMD_M2M, EXYNOS_DRM_DEGREE_0); if (ret) { fprintf(stderr, "failed to ipp property\n"); return; } for (i = 0; i < MAX_BUF; i++) { /* For source buffer map to IPP */ ret = exynos_drm_ipp_queue_buf(fd, &qbuf1[i], EXYNOS_DRM_OPS_SRC, IPP_BUF_ENQUEUE, property.prop_id, i, gem1[i].handle); if (ret) { fprintf(stderr, "failed to ipp buf src map\n"); goto err_ipp_ctrl_close; } } for (i = 0; i < MAX_BUF; i++) { /* For destination buffer map to IPP */ ret = exynos_drm_ipp_queue_buf(fd, &qbuf2[i], EXYNOS_DRM_OPS_DST, IPP_BUF_ENQUEUE, property.prop_id, i, gem2[i].handle); if (ret) { fprintf(stderr, "failed to ipp buf dst map\n"); goto err_ipp_ctrl_close; } } /* Start */ gettimeofday(&begin, NULL); ret = exynos_drm_ipp_cmd_ctrl(fd, &cmd_ctrl, property.prop_id, IPP_CTRL_PLAY); if (ret) { fprintf(stderr, "failed to ipp ctrl IPP_CMD_M2M start\n"); goto err_ipp_ctrl_close; } j=0; while (1) { struct timeval timeout = { .tv_sec = 3, .tv_usec = 0 }; fd_set fds; FD_ZERO(&fds); FD_SET(0, &fds); FD_SET(fd, &fds); ret = select(fd + 1, &fds, NULL, NULL, &timeout); if (ret <= 0) { fprintf(stderr, "select timed out or error.\n"); continue; } else if (FD_ISSET(0, &fds)) { break; } gettimeofday(&end, NULL); usec[j] = (end.tv_sec - begin.tv_sec) * 1000000 + (end.tv_usec - begin.tv_usec); if(ipp_event_handler() < 0) break; if (++j > MAX_LOOP) break; gettimeofday(&begin, NULL); } err_ipp_ctrl_close: /* For source buffer dequeue to IPP */ for (i = 0; i < MAX_BUF; i++) { ret = exynos_drm_ipp_queue_buf(fd, &qbuf1[i], EXYNOS_DRM_OPS_SRC, IPP_BUF_DEQUEUE, property.prop_id, i, gem1[i].handle); if (ret < 0) fprintf(stderr, "failed to ipp buf dst dequeue\n"); } /* For destination buffer dequeue to IPP */ for (i = 0; i < MAX_BUF; i++) { ret = exynos_drm_ipp_queue_buf(fd, &qbuf2[i], EXYNOS_DRM_OPS_DST, IPP_BUF_DEQUEUE, property.prop_id, i, gem2[i].handle); if (ret < 0) fprintf(stderr, "failed to ipp buf dst dequeue\n"); } /* Stop */ ret = exynos_drm_ipp_cmd_ctrl(fd, &cmd_ctrl, property.prop_id, IPP_CTRL_STOP); if (ret) fprintf(stderr, "failed to ipp ctrl IPP_CMD_WB stop\n"); return; } Eunchul Kim (4): drm/exynos: add ipp subsystem drm/exynos: add fimc ipp driver drm/exynos: add rotator ipp driver drm/exynos: add gsc ipp driver drivers/gpu/drm/exynos/Kconfig | 24 + drivers/gpu/drm/exynos/Makefile | 4 + drivers/gpu/drm/exynos/exynos_drm_drv.c | 69 + drivers/gpu/drm/exynos/exynos_drm_drv.h | 10 + drivers/gpu/drm/exynos/exynos_drm_fimc.c | 2041 +++++++++++++++++++++++++++ drivers/gpu/drm/exynos/exynos_drm_fimc.h | 35 + drivers/gpu/drm/exynos/exynos_drm_gsc.c | 1476 +++++++++++++++++++ drivers/gpu/drm/exynos/exynos_drm_gsc.h | 35 + drivers/gpu/drm/exynos/exynos_drm_ipp.c | 1931 +++++++++++++++++++++++++ drivers/gpu/drm/exynos/exynos_drm_ipp.h | 268 ++++ drivers/gpu/drm/exynos/exynos_drm_rotator.c | 801 +++++++++++ drivers/gpu/drm/exynos/exynos_drm_rotator.h | 13 + drivers/gpu/drm/exynos/regs-fimc.h | 669 +++++++++ drivers/gpu/drm/exynos/regs-gsc.h | 295 ++++ drivers/gpu/drm/exynos/regs-rotator.h | 73 + include/drm/exynos_drm.h | 33 + include/uapi/drm/exynos_drm.h | 189 +++ 17 files changed, 7966 insertions(+), 0 deletions(-) create mode 100644 drivers/gpu/drm/exynos/exynos_drm_fimc.c create mode 100644 drivers/gpu/drm/exynos/exynos_drm_fimc.h create mode 100644 drivers/gpu/drm/exynos/exynos_drm_gsc.c create mode 100644 drivers/gpu/drm/exynos/exynos_drm_gsc.h create mode 100644 drivers/gpu/drm/exynos/exynos_drm_ipp.c create mode 100644 drivers/gpu/drm/exynos/exynos_drm_ipp.h create mode 100644 drivers/gpu/drm/exynos/exynos_drm_rotator.c create mode 100644 drivers/gpu/drm/exynos/exynos_drm_rotator.h create mode 100644 drivers/gpu/drm/exynos/regs-fimc.h create mode 100644 drivers/gpu/drm/exynos/regs-gsc.h create mode 100644 drivers/gpu/drm/exynos/regs-rotator.h

3 8

[PATCH 1/3] drm/exynos: modify wait for vsync functions to use wait queues
by Prathyush K 04 Dec '12

04 Dec '12

It is more optimium to use wait queues while waiting for vsync so that the current task is put to sleep. This way, the task wont hog the CPU while waiting. We use wait_event_timeout and not an interruptible function since we dont want the function to exit when a signal is pending (e.g. drm release). This patch modifies the wait for vblank functions of both fimd and mixer. Also, the current fimd wait_for_vblank did not work in exynos5 since VIDCON1 has to be used in addition to timing base offset (0x20000). This patch also eliminates this problem. Signed-off-by: Prathyush K <prathyush.k(a)samsung.com> --- drivers/gpu/drm/exynos/exynos_drm_drv.h | 14 -------------- drivers/gpu/drm/exynos/exynos_drm_fimd.c | 21 +++++++++++++++++---- drivers/gpu/drm/exynos/exynos_mixer.c | 21 ++++++++++++++++----- 3 files changed, 33 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.h b/drivers/gpu/drm/exynos/exynos_drm_drv.h index a4702a8..6bb8644 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_drv.h +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.h @@ -36,20 +36,6 @@ #define MAX_FB_BUFFER 4 #define DEFAULT_ZPOS -1 -#define _wait_for(COND, MS) ({ \ - unsigned long timeout__ = jiffies + msecs_to_jiffies(MS); \ - int ret__ = 0; \ - while (!(COND)) { \ - if (time_after(jiffies, timeout__)) { \ - ret__ = -ETIMEDOUT; \ - break; \ - } \ - } \ - ret__; \ -}) - -#define wait_for(COND, MS) _wait_for(COND, MS) - struct drm_device; struct exynos_drm_overlay; struct drm_connector; diff --git a/drivers/gpu/drm/exynos/exynos_drm_fimd.c b/drivers/gpu/drm/exynos/exynos_drm_fimd.c index e2abae6..63f0d0f 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_fimd.c +++ b/drivers/gpu/drm/exynos/exynos_drm_fimd.c @@ -101,6 +101,8 @@ struct fimd_context { u32 vidcon1; bool suspended; struct mutex lock; + wait_queue_head_t wait_vsync_queue; + atomic_t wait_vsync_event; struct exynos_drm_panel_info *panel; }; @@ -606,11 +608,16 @@ static void fimd_win_disable(struct device *dev, int zpos) static void fimd_wait_for_vblank(struct device *dev) { struct fimd_context *ctx = get_fimd_context(dev); - int ret; - ret = wait_for((__raw_readl(ctx->regs + VIDCON1) & - VIDCON1_VSTATUS_VSYNC), 50); - if (ret < 0) + atomic_set(&ctx->wait_vsync_event, 1); + + /* + * wait for FIMD to signal VSYNC interrupt or return after + * timeout which is set to 50ms (refresh rate of 20). + */ + if (!wait_event_timeout(ctx->wait_vsync_queue, + !atomic_read(&ctx->wait_vsync_event), + DRM_HZ/20)) DRM_DEBUG_KMS("vblank wait timed out.\n"); } @@ -677,6 +684,10 @@ static irqreturn_t fimd_irq_handler(int irq, void *dev_id) drm_handle_vblank(drm_dev, manager->pipe); fimd_finish_pageflip(drm_dev, manager->pipe); + /* set wait vsync event to zero and wake up queue. */ + atomic_set(&ctx->wait_vsync_event, 0); + DRM_WAKEUP(&ctx->wait_vsync_queue); + out: return IRQ_HANDLED; } @@ -967,6 +978,8 @@ static int __devinit fimd_probe(struct platform_device *pdev) ctx->vidcon1 = pdata->vidcon1; ctx->default_win = pdata->default_win; ctx->panel = panel; + DRM_INIT_WAITQUEUE(&ctx->wait_vsync_queue); + atomic_set(&ctx->wait_vsync_event, 0); subdrv = &ctx->subdrv; diff --git a/drivers/gpu/drm/exynos/exynos_mixer.c b/drivers/gpu/drm/exynos/exynos_mixer.c index 50100cf..972281e 100644 --- a/drivers/gpu/drm/exynos/exynos_mixer.c +++ b/drivers/gpu/drm/exynos/exynos_mixer.c @@ -93,6 +93,8 @@ struct mixer_context { struct hdmi_win_data win_data[MIXER_WIN_NR]; enum mixer_version_id mxr_ver; void *parent_ctx; + wait_queue_head_t wait_vsync_queue; + atomic_t wait_vsync_event; }; struct mixer_drv_data { @@ -791,12 +793,16 @@ static void mixer_dpms(void *ctx, int mode) static void mixer_wait_for_vblank(void *ctx) { struct mixer_context *mixer_ctx = ctx; - struct mixer_resources *res = &mixer_ctx->mixer_res; - int ret; - ret = wait_for((mixer_reg_read(res, MXR_INT_STATUS) & - MXR_INT_STATUS_VSYNC), 50); - if (ret < 0) + atomic_set(&mixer_ctx->wait_vsync_event, 1); + + /* + * wait for MIXER to signal VSYNC interrupt or return after + * timeout which is set to 50ms (refresh rate of 20). + */ + if (!wait_event_timeout(mixer_ctx->wait_vsync_queue, + !atomic_read(&mixer_ctx->wait_vsync_event), + DRM_HZ/20)) DRM_DEBUG_KMS("vblank wait timed out.\n"); } @@ -957,6 +963,9 @@ static irqreturn_t mixer_irq_handler(int irq, void *arg) drm_handle_vblank(drm_hdmi_ctx->drm_dev, ctx->pipe); mixer_finish_pageflip(drm_hdmi_ctx->drm_dev, ctx->pipe); + + atomic_set(&ctx->wait_vsync_event, 0); + DRM_WAKEUP(&ctx->wait_vsync_queue); } out: @@ -1166,6 +1175,8 @@ static int __devinit mixer_probe(struct platform_device *pdev) drm_hdmi_ctx->ctx = (void *)ctx; ctx->vp_enabled = drv->is_vp_enabled; ctx->mxr_ver = drv->version; + DRM_INIT_WAITQUEUE(&ctx->wait_vsync_queue); + atomic_set(&ctx->wait_vsync_event, 0); platform_set_drvdata(pdev, drm_hdmi_ctx); -- 1.7.0.4

1 2

[PATCH] drm/exynos: enable iommu for mixer and not hdmi
by Prathyush K 04 Dec '12

04 Dec '12

According to the new IOMMU framework for exynos sysmmus, the owner of the sysmmu-tv is mixer (which is the actual device that does DMA) and not hdmi. The mmu-master in sysmmu-tv node is set as below in exynos5250.dtsi. sysmmu-tv { - mmu-master = <&mixer>; }; This patch moves the iommu_on function from the hdmi context to the mixer context. And this patch is based on exynos-drm-next-iommu branch of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos Signed-off-by: Prathyush K <prathyush.k(a)samsung.com> --- drivers/gpu/drm/exynos/exynos_drm_hdmi.c | 8 ++++---- drivers/gpu/drm/exynos/exynos_drm_hdmi.h | 2 +- drivers/gpu/drm/exynos/exynos_hdmi.c | 24 ------------------------ drivers/gpu/drm/exynos/exynos_mixer.c | 23 +++++++++++++++++++++++ 4 files changed, 28 insertions(+), 29 deletions(-) diff --git a/drivers/gpu/drm/exynos/exynos_drm_hdmi.c b/drivers/gpu/drm/exynos/exynos_drm_hdmi.c index 997fb6e..8b771a3 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_hdmi.c +++ b/drivers/gpu/drm/exynos/exynos_drm_hdmi.c @@ -368,8 +368,8 @@ static int hdmi_subdrv_probe(struct drm_device *drm_dev, ctx->hdmi_ctx->drm_dev = drm_dev; ctx->mixer_ctx->drm_dev = drm_dev; - if (hdmi_ops->iommu_on) - hdmi_ops->iommu_on(ctx->hdmi_ctx->ctx, true); + if (mixer_ops->iommu_on) + mixer_ops->iommu_on(ctx->mixer_ctx->ctx, true); return 0; } @@ -381,8 +381,8 @@ static void hdmi_subdrv_remove(struct drm_device *drm_dev, struct device *dev) ctx = get_ctx_from_subdrv(subdrv); - if (hdmi_ops->iommu_on) - hdmi_ops->iommu_on(ctx->hdmi_ctx->ctx, false); + if (mixer_ops->iommu_on) + mixer_ops->iommu_on(ctx->mixer_ctx->ctx, false); } static int __devinit exynos_drm_hdmi_probe(struct platform_device *pdev) diff --git a/drivers/gpu/drm/exynos/exynos_drm_hdmi.h b/drivers/gpu/drm/exynos/exynos_drm_hdmi.h index 5c033d1..54b5223 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_hdmi.h +++ b/drivers/gpu/drm/exynos/exynos_drm_hdmi.h @@ -50,7 +50,6 @@ struct exynos_hdmi_ops { int (*power_on)(void *ctx, int mode); /* manager */ - int (*iommu_on)(void *ctx, bool enable); void (*mode_fixup)(void *ctx, struct drm_connector *connector, const struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode); @@ -63,6 +62,7 @@ struct exynos_hdmi_ops { struct exynos_mixer_ops { /* manager */ + int (*iommu_on)(void *ctx, bool enable); int (*enable_vblank)(void *ctx, int pipe); void (*disable_vblank)(void *ctx); void (*dpms)(void *ctx, int mode); diff --git a/drivers/gpu/drm/exynos/exynos_hdmi.c b/drivers/gpu/drm/exynos/exynos_hdmi.c index d1a1d71..7df1d85 100644 --- a/drivers/gpu/drm/exynos/exynos_hdmi.c +++ b/drivers/gpu/drm/exynos/exynos_hdmi.c @@ -40,7 +40,6 @@ #include "exynos_drm_drv.h" #include "exynos_drm_hdmi.h" -#include "exynos_drm_iommu.h" #include "exynos_hdmi.h" @@ -68,7 +67,6 @@ struct hdmi_resources { struct hdmi_context { struct device *dev; - struct drm_device *drm_dev; bool hpd; bool powered; bool dvi_mode; @@ -85,7 +83,6 @@ struct hdmi_context { int cur_conf; struct hdmi_resources res; - void *parent_ctx; int hpd_gpio; @@ -1947,25 +1944,6 @@ static void hdmi_conf_apply(struct hdmi_context *hdata) hdmi_regs_dump(hdata, "start"); } -static int hdmi_iommu_on(void *ctx, bool enable) -{ - struct exynos_drm_hdmi_context *drm_hdmi_ctx; - struct hdmi_context *hdata = ctx; - struct drm_device *drm_dev; - - drm_hdmi_ctx = hdata->parent_ctx; - drm_dev = drm_hdmi_ctx->drm_dev; - - if (is_drm_iommu_supported(drm_dev)) { - if (enable) - return drm_iommu_attach_device(drm_dev, hdata->dev); - - drm_iommu_detach_device(drm_dev, hdata->dev); - } - - return 0; -} - static void hdmi_mode_fixup(void *ctx, struct drm_connector *connector, const struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) @@ -2122,7 +2100,6 @@ static struct exynos_hdmi_ops hdmi_ops = { .check_timing = hdmi_check_timing, /* manager */ - .iommu_on = hdmi_iommu_on, .mode_fixup = hdmi_mode_fixup, .mode_set = hdmi_mode_set, .get_max_resol = hdmi_get_max_resol, @@ -2379,7 +2356,6 @@ static int __devinit hdmi_probe(struct platform_device *pdev) mutex_init(&hdata->hdmi_mutex); drm_hdmi_ctx->ctx = (void *)hdata; - hdata->parent_ctx = (void *)drm_hdmi_ctx; platform_set_drvdata(pdev, drm_hdmi_ctx); diff --git a/drivers/gpu/drm/exynos/exynos_mixer.c b/drivers/gpu/drm/exynos/exynos_mixer.c index 614b2e9..574ed3f 100644 --- a/drivers/gpu/drm/exynos/exynos_mixer.c +++ b/drivers/gpu/drm/exynos/exynos_mixer.c @@ -36,6 +36,7 @@ #include "exynos_drm_drv.h" #include "exynos_drm_hdmi.h" +#include "exynos_drm_iommu.h" #define get_mixer_context(dev) platform_get_drvdata(to_platform_device(dev)) @@ -80,6 +81,7 @@ enum mixer_version_id { struct mixer_context { struct device *dev; + struct drm_device *drm_dev; int pipe; bool interlace; bool powered; @@ -90,6 +92,7 @@ struct mixer_context { struct mixer_resources mixer_res; struct hdmi_win_data win_data[MIXER_WIN_NR]; enum mixer_version_id mxr_ver; + void *parent_ctx; }; struct mixer_drv_data { @@ -665,6 +668,24 @@ static void mixer_win_reset(struct mixer_context *ctx) spin_unlock_irqrestore(&res->reg_slock, flags); } +static int mixer_iommu_on(void *ctx, bool enable) +{ + struct exynos_drm_hdmi_context *drm_hdmi_ctx; + struct mixer_context *mdata = ctx; + struct drm_device *drm_dev; + + drm_hdmi_ctx = mdata->parent_ctx; + drm_dev = drm_hdmi_ctx->drm_dev; + + if (is_drm_iommu_supported(drm_dev)) { + if (enable) + return drm_iommu_attach_device(drm_dev, mdata->dev); + + drm_iommu_detach_device(drm_dev, mdata->dev); + } + return 0; +} + static void mixer_poweron(struct mixer_context *ctx) { struct mixer_resources *res = &ctx->mixer_res; @@ -866,6 +887,7 @@ static void mixer_win_disable(void *ctx, int win) static struct exynos_mixer_ops mixer_ops = { /* manager */ + .iommu_on = mixer_iommu_on, .enable_vblank = mixer_enable_vblank, .disable_vblank = mixer_disable_vblank, .dpms = mixer_dpms, @@ -1149,6 +1171,7 @@ static int __devinit mixer_probe(struct platform_device *pdev) } ctx->dev = &pdev->dev; + ctx->parent_ctx = (void *)drm_hdmi_ctx; drm_hdmi_ctx->ctx = (void *)ctx; ctx->vp_enabled = drv->is_vp_enabled; ctx->mxr_ver = drv->version; -- 1.7.0.4

2 2

[PATCH 0/3] drm/exynos: add iommu support for -next
by Inki Dae 04 Dec '12

04 Dec '12

Hi all, This patch set adds iommu support for exynos drm framework and also relevant drivers, fimd and hdmi. Exynos4xxx SoC and later have some iommu hardware units for video codec, Camera, Post Processer(FIMC and GScaler), FIMD, HDMI, Graphics 2D and so on. In other words, each device has its own iommu hardware unit. So each device could have its own device address space. But in case of using DRM GEM and DMA Mapping API as allocator, there is one issue. the issue is that we can't know which device a gem buffer should be allocated for. So this patch set will use unified device address space. Simply saying, exynos drm-based all devices of using iommu have same device address space. For example, if user process requested gem allocation, exynos drm gem framework allocates desired physical memory region and maps it with unified iommu mapping table through DMA Mapping API call, dma_allc_attrs function. And this patch set is based on the following patch, "DMA-mapping & IOMMU - physically contiguous allocations" Marek posted before as RFC. For this, you can refer to below link, http://www.serverphorums.com/read.php?12,581741 As you know, RFC to the above patch is in progress so we have to wait for the patch to be merged to mainline. Thanks, Inki Dae Inki Dae (3): drm/exynos: add iommu support for exynos drm framework drm/exynos: add iommu support to fimd driver drm/exynos: add iommu support for hdmi driver drivers/gpu/drm/exynos/Kconfig | 6 + drivers/gpu/drm/exynos/Makefile | 1 + drivers/gpu/drm/exynos/exynos_drm_buf.c | 88 +++++------- drivers/gpu/drm/exynos/exynos_drm_dmabuf.c | 87 +++++------- drivers/gpu/drm/exynos/exynos_drm_drv.c | 23 +++- drivers/gpu/drm/exynos/exynos_drm_drv.h | 11 ++ drivers/gpu/drm/exynos/exynos_drm_fb.c | 52 ++++++- drivers/gpu/drm/exynos/exynos_drm_fimd.c | 9 +- drivers/gpu/drm/exynos/exynos_drm_gem.c | 210 +++-------------------------- drivers/gpu/drm/exynos/exynos_drm_gem.h | 1 + drivers/gpu/drm/exynos/exynos_drm_hdmi.c | 15 +++ drivers/gpu/drm/exynos/exynos_drm_hdmi.h | 1 + drivers/gpu/drm/exynos/exynos_drm_iommu.c | 156 +++++++++++++++++++++ drivers/gpu/drm/exynos/exynos_drm_iommu.h | 85 ++++++++++++ drivers/gpu/drm/exynos/exynos_hdmi.c | 21 +++ 15 files changed, 460 insertions(+), 306 deletions(-) create mode 100644 drivers/gpu/drm/exynos/exynos_drm_iommu.c create mode 100644 drivers/gpu/drm/exynos/exynos_drm_iommu.h -- 1.8.0.rc3.16.g8ead1bf

2 7

[PATCH v2] drm/exynos: create new sgt from existing sgt
by Prathyush K 04 Dec '12

04 Dec '12

Changelog v2: Implement copy sgt code in exynos_get_sgt itself instead of calling sg_clone_table. Changelog v1: During map_dma_buf, a new sgt needs to be created before being mapped to another device's address space. Currently, this sgt is created by calling dma_get_sgtable which creates a sgt from pages. This will be time consuming if the map/unmap calls are done repeatedly for very large buffers. Instead, we can just use the sgt which was already created during buffer allocation and create the new sgt by copying the sgls from the existing sgt. And this patch is based on exynos-drm-next-iommu branch of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos Signed-off-by: Prathyush K <prathyush.k(a)samsung.com> --- drivers/gpu/drm/exynos/exynos_drm_dmabuf.c | 18 +++++++++++++++--- 1 files changed, 15 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c b/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c index d9307bd..91c7db1 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c +++ b/drivers/gpu/drm/exynos/exynos_drm_dmabuf.c @@ -34,19 +34,31 @@ static struct sg_table *exynos_get_sgt(struct drm_device *drm_dev, struct exynos_drm_gem_buf *buf) { struct sg_table *sgt = NULL; + struct scatterlist *s, *s_out; + unsigned int i; int ret; + if (!buf->sgt) { + DRM_ERROR("no sg table.\n"); + goto out; + } + sgt = kzalloc(sizeof(*sgt), GFP_KERNEL); if (!sgt) goto out; - ret = dma_get_sgtable(drm_dev->dev, sgt, buf->kvaddr, - buf->dma_addr, buf->size); + ret = sg_alloc_table(sgt, buf->sgt->orig_nents, GFP_KERNEL); if (ret < 0) { - DRM_ERROR("failed to get sgtable.\n"); + DRM_ERROR("failed to alloc table\n"); goto err_free_sgt; } + s_out = sgt->sgl; + for_each_sg(buf->sgt->sgl, s, buf->sgt->orig_nents, i) { + sg_set_page(s_out, sg_page(s), s->length, s->offset); + s_out = sg_next(s_out); + } + return sgt; err_free_sgt: -- 1.7.0.4

2 1

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

dri-devel December 2012