From: Christian König deathsimple@vodafone.de

Different rings have different criteria to test if they are stuck.

v2: rebased on current drm-next

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 4 +- drivers/gpu/drm/radeon/radeon_asic.c | 44 ++++++++++++++++++-------------- drivers/gpu/drm/radeon/radeon_fence.c | 2 +- 3 files changed, 28 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 138b952..bea99e3 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1144,7 +1144,6 @@ struct radeon_asic { int (*resume)(struct radeon_device *rdev); int (*suspend)(struct radeon_device *rdev); void (*vga_set_state)(struct radeon_device *rdev, bool state); - bool (*gpu_is_lockup)(struct radeon_device *rdev, struct radeon_ring *cp); int (*asic_reset)(struct radeon_device *rdev); /* ioctl hw specific callback. Some hw might want to perform special * operation on specific ioctl. For instance on wait idle some hw @@ -1173,6 +1172,7 @@ struct radeon_asic { void (*ring_start)(struct radeon_device *rdev, struct radeon_ring *cp); int (*ring_test)(struct radeon_device *rdev, struct radeon_ring *cp); int (*ib_test)(struct radeon_device *rdev, struct radeon_ring *cp); + bool (*is_lockup)(struct radeon_device *rdev, struct radeon_ring *cp); } ring[RADEON_NUM_RINGS]; /* irqs */ struct { @@ -1730,7 +1730,6 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_suspend(rdev) (rdev)->asic->suspend((rdev)) #define radeon_cs_parse(rdev, r, p) (rdev)->asic->ring[(r)].cs_parse((p)) #define radeon_vga_set_state(rdev, state) (rdev)->asic->vga_set_state((rdev), (state)) -#define radeon_gpu_is_lockup(rdev, cp) (rdev)->asic->gpu_is_lockup((rdev), (cp)) #define radeon_asic_reset(rdev) (rdev)->asic->asic_reset((rdev)) #define radeon_gart_tlb_flush(rdev) (rdev)->asic->gart.tlb_flush((rdev)) #define radeon_gart_set_page(rdev, i, p) (rdev)->asic->gart.set_page((rdev), (i), (p)) @@ -1739,6 +1738,7 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_ib_test(rdev, r, cp) (rdev)->asic->ring[(r)].ib_test((rdev), (cp)) #define radeon_ring_ib_execute(rdev, r, ib) (rdev)->asic->ring[(r)].ib_execute((rdev), (ib)) #define radeon_ring_ib_parse(rdev, r, ib) (rdev)->asic->ring[(r)].ib_parse((rdev), (ib)) +#define radeon_ring_is_lockup(rdev, r, cp) (rdev)->asic->ring[(r)].is_lockup((rdev), (cp)) #define radeon_irq_set(rdev) (rdev)->asic->irq.set((rdev)) #define radeon_irq_process(rdev) (rdev)->asic->irq.process((rdev)) #define radeon_get_vblank_counter(rdev, crtc) (rdev)->asic->display.get_vblank_counter((rdev), (crtc)) diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index be4dc2f..958b9ea 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -134,7 +134,6 @@ static struct radeon_asic r100_asic = { .suspend = &r100_suspend, .resume = &r100_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r100_gpu_is_lockup, .asic_reset = &r100_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -152,6 +151,7 @@ static struct radeon_asic r100_asic = { .ring_start = &r100_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -208,7 +208,6 @@ static struct radeon_asic r200_asic = { .suspend = &r100_suspend, .resume = &r100_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r100_gpu_is_lockup, .asic_reset = &r100_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -226,6 +225,7 @@ static struct radeon_asic r200_asic = { .ring_start = &r100_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -282,7 +282,6 @@ static struct radeon_asic r300_asic = { .suspend = &r300_suspend, .resume = &r300_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &r300_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -300,6 +299,7 @@ static struct radeon_asic r300_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -356,7 +356,6 @@ static struct radeon_asic r300_asic_pcie = { .suspend = &r300_suspend, .resume = &r300_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &r300_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -374,6 +373,7 @@ static struct radeon_asic r300_asic_pcie = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -430,7 +430,6 @@ static struct radeon_asic r420_asic = { .suspend = &r420_suspend, .resume = &r420_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &r300_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -448,6 +447,7 @@ static struct radeon_asic r420_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -504,7 +504,6 @@ static struct radeon_asic rs400_asic = { .suspend = &rs400_suspend, .resume = &rs400_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &r300_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -522,6 +521,7 @@ static struct radeon_asic rs400_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -578,7 +578,6 @@ static struct radeon_asic rs600_asic = { .suspend = &rs600_suspend, .resume = &rs600_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &rs600_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -596,6 +595,7 @@ static struct radeon_asic rs600_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -652,7 +652,6 @@ static struct radeon_asic rs690_asic = { .suspend = &rs690_suspend, .resume = &rs690_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &rs600_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -670,6 +669,7 @@ static struct radeon_asic rs690_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -726,7 +726,6 @@ static struct radeon_asic rv515_asic = { .suspend = &rv515_suspend, .resume = &rv515_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &rs600_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -744,6 +743,7 @@ static struct radeon_asic rv515_asic = { .ring_start = &rv515_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -800,7 +800,6 @@ static struct radeon_asic r520_asic = { .suspend = &rv515_suspend, .resume = &r520_resume, .vga_set_state = &r100_vga_set_state, - .gpu_is_lockup = &r300_gpu_is_lockup, .asic_reset = &rs600_asic_reset, .ioctl_wait_idle = NULL, .gui_idle = &r100_gui_idle, @@ -818,6 +817,7 @@ static struct radeon_asic r520_asic = { .ring_start = &rv515_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, + .is_lockup = &r300_gpu_is_lockup, } }, .irq = { @@ -874,7 +874,6 @@ static struct radeon_asic r600_asic = { .suspend = &r600_suspend, .resume = &r600_resume, .vga_set_state = &r600_vga_set_state, - .gpu_is_lockup = &r600_gpu_is_lockup, .asic_reset = &r600_asic_reset, .ioctl_wait_idle = r600_ioctl_wait_idle, .gui_idle = &r600_gui_idle, @@ -891,6 +890,7 @@ static struct radeon_asic r600_asic = { .cs_parse = &r600_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &r600_gpu_is_lockup, } }, .irq = { @@ -946,7 +946,6 @@ static struct radeon_asic rs780_asic = { .fini = &r600_fini, .suspend = &r600_suspend, .resume = &r600_resume, - .gpu_is_lockup = &r600_gpu_is_lockup, .vga_set_state = &r600_vga_set_state, .asic_reset = &r600_asic_reset, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -964,6 +963,7 @@ static struct radeon_asic rs780_asic = { .cs_parse = &r600_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &r600_gpu_is_lockup, } }, .irq = { @@ -1020,7 +1020,6 @@ static struct radeon_asic rv770_asic = { .suspend = &rv770_suspend, .resume = &rv770_resume, .asic_reset = &r600_asic_reset, - .gpu_is_lockup = &r600_gpu_is_lockup, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, .gui_idle = &r600_gui_idle, @@ -1037,6 +1036,7 @@ static struct radeon_asic rv770_asic = { .cs_parse = &r600_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &r600_gpu_is_lockup, } }, .irq = { @@ -1092,7 +1092,6 @@ static struct radeon_asic evergreen_asic = { .fini = &evergreen_fini, .suspend = &evergreen_suspend, .resume = &evergreen_resume, - .gpu_is_lockup = &evergreen_gpu_is_lockup, .asic_reset = &evergreen_asic_reset, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -1110,6 +1109,7 @@ static struct radeon_asic evergreen_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &evergreen_gpu_is_lockup, } }, .irq = { @@ -1165,7 +1165,6 @@ static struct radeon_asic sumo_asic = { .fini = &evergreen_fini, .suspend = &evergreen_suspend, .resume = &evergreen_resume, - .gpu_is_lockup = &evergreen_gpu_is_lockup, .asic_reset = &evergreen_asic_reset, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -1183,6 +1182,7 @@ static struct radeon_asic sumo_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &evergreen_gpu_is_lockup, }, }, .irq = { @@ -1238,7 +1238,6 @@ static struct radeon_asic btc_asic = { .fini = &evergreen_fini, .suspend = &evergreen_suspend, .resume = &evergreen_resume, - .gpu_is_lockup = &evergreen_gpu_is_lockup, .asic_reset = &evergreen_asic_reset, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -1256,6 +1255,7 @@ static struct radeon_asic btc_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &evergreen_gpu_is_lockup, } }, .irq = { @@ -1321,7 +1321,6 @@ static struct radeon_asic cayman_asic = { .fini = &cayman_fini, .suspend = &cayman_suspend, .resume = &cayman_resume, - .gpu_is_lockup = &cayman_gpu_is_lockup, .asic_reset = &cayman_asic_reset, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -1340,6 +1339,7 @@ static struct radeon_asic cayman_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &cayman_gpu_is_lockup, }, [CAYMAN_RING_TYPE_CP1_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1349,6 +1349,7 @@ static struct radeon_asic cayman_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &cayman_gpu_is_lockup, }, [CAYMAN_RING_TYPE_CP2_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1358,6 +1359,7 @@ static struct radeon_asic cayman_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &cayman_gpu_is_lockup, } }, .irq = { @@ -1413,7 +1415,6 @@ static struct radeon_asic trinity_asic = { .fini = &cayman_fini, .suspend = &cayman_suspend, .resume = &cayman_resume, - .gpu_is_lockup = &cayman_gpu_is_lockup, .asic_reset = &cayman_asic_reset, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -1432,6 +1433,7 @@ static struct radeon_asic trinity_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &cayman_gpu_is_lockup, }, [CAYMAN_RING_TYPE_CP1_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1441,6 +1443,7 @@ static struct radeon_asic trinity_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &cayman_gpu_is_lockup, }, [CAYMAN_RING_TYPE_CP2_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1450,6 +1453,7 @@ static struct radeon_asic trinity_asic = { .cs_parse = &evergreen_cs_parse, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &cayman_gpu_is_lockup, } }, .irq = { @@ -1515,7 +1519,6 @@ static struct radeon_asic si_asic = { .fini = &si_fini, .suspend = &si_suspend, .resume = &si_resume, - .gpu_is_lockup = &si_gpu_is_lockup, .asic_reset = &si_asic_reset, .vga_set_state = &r600_vga_set_state, .ioctl_wait_idle = r600_ioctl_wait_idle, @@ -1534,6 +1537,7 @@ static struct radeon_asic si_asic = { .cs_parse = NULL, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &si_gpu_is_lockup, }, [CAYMAN_RING_TYPE_CP1_INDEX] = { .ib_execute = &si_ring_ib_execute, @@ -1543,6 +1547,7 @@ static struct radeon_asic si_asic = { .cs_parse = NULL, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &si_gpu_is_lockup, }, [CAYMAN_RING_TYPE_CP2_INDEX] = { .ib_execute = &si_ring_ib_execute, @@ -1552,6 +1557,7 @@ static struct radeon_asic si_asic = { .cs_parse = NULL, .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, + .is_lockup = &si_gpu_is_lockup, } }, .irq = { diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 4bd36a3..66b2229 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -259,7 +259,7 @@ retry: * if we experiencing a lockup the value doesn't change */ if (seq == rdev->fence_drv[fence->ring].last_seq && - radeon_gpu_is_lockup(rdev, &rdev->ring[fence->ring])) { + radeon_ring_is_lockup(rdev, fence->ring, &rdev->ring[fence->ring])) { /* good news we believe it's a lockup */ printk(KERN_WARNING "GPU lockup (waiting for 0x%08X last fence id 0x%08X)\n", fence->seq, seq);

From: Christian König deathsimple@vodafone.de

Just register the debugfs files on init instead of checking the chipset type multiple times.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon_ring.c | 31 +++++++++++++++++++------------ 1 files changed, 19 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index cc33b3d..b6eb1d2 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -34,7 +34,7 @@ #include "atom.h"

int radeon_debugfs_ib_init(struct radeon_device *rdev); -int radeon_debugfs_ring_init(struct radeon_device *rdev); +int radeon_debugfs_ring_init(struct radeon_device *rdev, struct radeon_ring *ring);

u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx) { @@ -237,9 +237,6 @@ int radeon_ib_pool_init(struct radeon_device *rdev) if (radeon_debugfs_ib_init(rdev)) { DRM_ERROR("Failed to register debugfs file for IB !\n"); } - if (radeon_debugfs_ring_init(rdev)) { - DRM_ERROR("Failed to register debugfs file for rings !\n"); - } radeon_mutex_unlock(&rdev->ib_pool.mutex); return 0; } @@ -411,6 +408,9 @@ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsig } ring->ptr_mask = (ring->ring_size / 4) - 1; ring->ring_free_dw = ring->ring_size / 4; + if (radeon_debugfs_ring_init(rdev, ring)) { + DRM_ERROR("Failed to register debugfs file for rings !\n"); + } return 0; }

@@ -501,17 +501,24 @@ static char radeon_debugfs_ib_names[RADEON_IB_POOL_SIZE][32]; static unsigned radeon_debugfs_ib_idx[RADEON_IB_POOL_SIZE]; #endif

-int radeon_debugfs_ring_init(struct radeon_device *rdev) +int radeon_debugfs_ring_init(struct radeon_device *rdev, struct radeon_ring *ring) { #if defined(CONFIG_DEBUG_FS) - if (rdev->family >= CHIP_CAYMAN) - return radeon_debugfs_add_files(rdev, radeon_debugfs_ring_info_list, - ARRAY_SIZE(radeon_debugfs_ring_info_list)); - else - return radeon_debugfs_add_files(rdev, radeon_debugfs_ring_info_list, 1); -#else - return 0; + unsigned i; + for (i = 0; i < ARRAY_SIZE(radeon_debugfs_ring_info_list); ++i) { + struct drm_info_list *info = &radeon_debugfs_ring_info_list[i]; + int ridx = *(int*)radeon_debugfs_ring_info_list[i].data; + unsigned r; + + if (&rdev->ring[ridx] != ring) + continue; + + r = radeon_debugfs_add_files(rdev, info, 1); + if (r) + return r; + } #endif + return 0; }

int radeon_debugfs_ib_init(struct radeon_device *rdev)

From: Christian König deathsimple@vodafone.de

Previusly multiple rings could trigger multiple GPU resets at the same time.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 3 +- drivers/gpu/drm/radeon/radeon_fence.c | 150 +++++++++++++++++---------------- 2 files changed, 77 insertions(+), 76 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 8801657..85a3aa9 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -255,8 +255,7 @@ struct radeon_fence_driver { volatile uint32_t *cpu_addr; atomic_t seq; uint32_t last_seq; - unsigned long last_jiffies; - unsigned long last_timeout; + unsigned long last_activity; wait_queue_head_t queue; struct list_head created; struct list_head emitted; diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 36c411f..1a9765a 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -74,6 +74,10 @@ int radeon_fence_emit(struct radeon_device *rdev, struct radeon_fence *fence) radeon_fence_ring_emit(rdev, fence->ring, fence); trace_radeon_fence_emit(rdev->ddev, fence->seq); fence->emitted = true; + /* are we the first fence on a previusly idle ring? */ + if (list_empty(&rdev->fence_drv[fence->ring].emitted)) { + rdev->fence_drv[fence->ring].last_activity = jiffies; + } list_move_tail(&fence->list, &rdev->fence_drv[fence->ring].emitted); write_unlock_irqrestore(&rdev->fence_lock, irq_flags); return 0; @@ -85,34 +89,14 @@ static bool radeon_fence_poll_locked(struct radeon_device *rdev, int ring) struct list_head *i, *n; uint32_t seq; bool wake = false; - unsigned long cjiffies;

seq = radeon_fence_read(rdev, ring); - if (seq != rdev->fence_drv[ring].last_seq) { - rdev->fence_drv[ring].last_seq = seq; - rdev->fence_drv[ring].last_jiffies = jiffies; - rdev->fence_drv[ring].last_timeout = RADEON_FENCE_JIFFIES_TIMEOUT; - } else { - cjiffies = jiffies; - if (time_after(cjiffies, rdev->fence_drv[ring].last_jiffies)) { - cjiffies -= rdev->fence_drv[ring].last_jiffies; - if (time_after(rdev->fence_drv[ring].last_timeout, cjiffies)) { - /* update the timeout */ - rdev->fence_drv[ring].last_timeout -= cjiffies; - } else { - /* the 500ms timeout is elapsed we should test - * for GPU lockup - */ - rdev->fence_drv[ring].last_timeout = 1; - } - } else { - /* wrap around update last jiffies, we will just wait - * a little longer - */ - rdev->fence_drv[ring].last_jiffies = cjiffies; - } + if (seq == rdev->fence_drv[ring].last_seq) return false; - } + + rdev->fence_drv[ring].last_seq = seq; + rdev->fence_drv[ring].last_activity = jiffies; + n = NULL; list_for_each(i, &rdev->fence_drv[ring].emitted) { fence = list_entry(i, struct radeon_fence, list); @@ -207,66 +191,84 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr) struct radeon_device *rdev; unsigned long irq_flags, timeout; u32 seq; - int r; + int i, r; + bool signaled;

if (fence == NULL) { WARN(1, "Querying an invalid fence : %p !\n", fence); - return 0; + return -EINVAL; } + rdev = fence->rdev; - if (radeon_fence_signaled(fence)) { - return 0; - } - timeout = rdev->fence_drv[fence->ring].last_timeout; -retry: - /* save current sequence used to check for GPU lockup */ - seq = rdev->fence_drv[fence->ring].last_seq; - trace_radeon_fence_wait_begin(rdev->ddev, seq); - if (intr) { + signaled = radeon_fence_signaled(fence); + while (!signaled) { + read_lock_irqsave(&rdev->fence_lock, irq_flags); + timeout = jiffies - RADEON_FENCE_JIFFIES_TIMEOUT; + if (time_after(rdev->fence_drv[fence->ring].last_activity, timeout)) { + /* the normal case, timeout is somewhere before last_activity */ + timeout = rdev->fence_drv[fence->ring].last_activity - timeout; + } else { + /* either jiffies wrapped around, or no fence was signaled in the last 500ms + * anyway we will just wait for the minimum amount and then check for a lockup */ + timeout = 1; + } + /* save current sequence value used to check for GPU lockups */ + seq = rdev->fence_drv[fence->ring].last_seq; + read_unlock_irqrestore(&rdev->fence_lock, irq_flags); + + trace_radeon_fence_wait_begin(rdev->ddev, seq); radeon_irq_kms_sw_irq_get(rdev, fence->ring); - r = wait_event_interruptible_timeout(rdev->fence_drv[fence->ring].queue, - radeon_fence_signaled(fence), timeout); + if (intr) { + r = wait_event_interruptible_timeout( + rdev->fence_drv[fence->ring].queue, + (signaled = radeon_fence_signaled(fence)), timeout); + } else { + r = wait_event_timeout( + rdev->fence_drv[fence->ring].queue, + (signaled = radeon_fence_signaled(fence)), timeout); + } radeon_irq_kms_sw_irq_put(rdev, fence->ring); if (unlikely(r < 0)) { return r; } - } else { - radeon_irq_kms_sw_irq_get(rdev, fence->ring); - r = wait_event_timeout(rdev->fence_drv[fence->ring].queue, - radeon_fence_signaled(fence), timeout); - radeon_irq_kms_sw_irq_put(rdev, fence->ring); - } - trace_radeon_fence_wait_end(rdev->ddev, seq); - if (unlikely(!radeon_fence_signaled(fence))) { - /* we were interrupted for some reason and fence isn't - * isn't signaled yet, resume wait - */ - if (r) { - timeout = r; - goto retry; - } - /* don't protect read access to rdev->fence_drv[t].last_seq - * if we experiencing a lockup the value doesn't change - */ - if (seq == rdev->fence_drv[fence->ring].last_seq && - radeon_ring_is_lockup(rdev, fence->ring, &rdev->ring[fence->ring])) { - - /* good news we believe it's a lockup */ - printk(KERN_WARNING "GPU lockup (waiting for 0x%08X last fence id 0x%08X)\n", - fence->seq, seq); - - /* mark the ring as not ready any more */ - rdev->ring[fence->ring].ready = false; - r = radeon_gpu_reset(rdev); - if (r) - return r; + trace_radeon_fence_wait_end(rdev->ddev, seq); + + if (unlikely(!signaled)) { + /* we were interrupted for some reason and fence + * isn't signaled yet, resume waiting */ + if (r) { + continue; + } + + write_lock_irqsave(&rdev->fence_lock, irq_flags); + /* check if sequence value has changed since last_activity */ + if (seq != rdev->fence_drv[fence->ring].last_seq) { + write_unlock_irqrestore(&rdev->fence_lock, irq_flags); + continue; + } + + /* change sequence value on all rings, so nobody else things there is a lockup */ + for (i = 0; i < RADEON_NUM_RINGS; ++i) + rdev->fence_drv[i].last_seq -= 0x10000; + write_unlock_irqrestore(&rdev->fence_lock, irq_flags); + + if (radeon_ring_is_lockup(rdev, fence->ring, &rdev->ring[fence->ring])) { + + /* good news we believe it's a lockup */ + printk(KERN_WARNING "GPU lockup (waiting for 0x%08X last fence id 0x%08X)\n", + fence->seq, seq); + + /* mark the ring as not ready any more */ + rdev->ring[fence->ring].ready = false; + r = radeon_gpu_reset(rdev); + if (r) + return r; + + write_lock_irqsave(&rdev->fence_lock, irq_flags); + rdev->fence_drv[fence->ring].last_activity = jiffies; + write_unlock_irqrestore(&rdev->fence_lock, irq_flags); + } } - timeout = RADEON_FENCE_JIFFIES_TIMEOUT; - write_lock_irqsave(&rdev->fence_lock, irq_flags); - rdev->fence_drv[fence->ring].last_timeout = RADEON_FENCE_JIFFIES_TIMEOUT; - rdev->fence_drv[fence->ring].last_jiffies = jiffies; - write_unlock_irqrestore(&rdev->fence_lock, irq_flags); - goto retry; } return 0; }

From: Christian König deathsimple@vodafone.de

Make the suballocator self containing to locking.

v2: split the bugfix into a seperate patch. v3: Jerome Glisse use mutex, no reason to use spinlock that are more heavyweight than mutex

Signed-off-by: Christian König deathsimple@vodafone.de Signed-off-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 1 + drivers/gpu/drm/radeon/radeon_sa.c | 15 +++++++++------ 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 85a3aa9..34c1041 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -381,6 +381,7 @@ struct radeon_bo_list { * alignment). */ struct radeon_sa_manager { + struct mutex mutex; struct radeon_bo *bo; struct list_head sa_bo; unsigned size; diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c index 8fbfe69..8c76c2e 100644 --- a/drivers/gpu/drm/radeon/radeon_sa.c +++ b/drivers/gpu/drm/radeon/radeon_sa.c @@ -37,6 +37,7 @@ int radeon_sa_bo_manager_init(struct radeon_device *rdev, { int r;

+ mutex_init(&sa_manager->mutex); sa_manager->bo = NULL; sa_manager->size = size; sa_manager->domain = domain; @@ -139,15 +140,15 @@ int radeon_sa_bo_new(struct radeon_device *rdev,

BUG_ON(align > RADEON_GPU_PAGE_SIZE); BUG_ON(size > sa_manager->size); + mutex_lock(&sa_manager->mutex);

/* no one ? */ - head = sa_manager->sa_bo.prev; if (list_empty(&sa_manager->sa_bo)) { + head = &sa_manager->sa_bo; goto out; }

/* look for a hole big enough */ - offset = 0; list_for_each_entry(tmp, &sa_manager->sa_bo, list) { /* room before this object ? */ if (offset < tmp->offset && (tmp->offset - offset) >= size) { @@ -157,9 +158,8 @@ int radeon_sa_bo_new(struct radeon_device *rdev, offset = tmp->offset + tmp->size; wasted = offset % align; if (wasted) { - wasted = align - wasted; + offset += align - wasted; } - offset += wasted; } /* room at the end ? */ head = sa_manager->sa_bo.prev; @@ -167,11 +167,11 @@ int radeon_sa_bo_new(struct radeon_device *rdev, offset = tmp->offset + tmp->size; wasted = offset % align; if (wasted) { - wasted = align - wasted; + offset += wasted = align - wasted; } - offset += wasted; if ((sa_manager->size - offset) < size) { /* failed to find somethings big enough */ + mutex_unlock(&sa_manager->mutex); return -ENOMEM; }

@@ -180,10 +180,13 @@ out: sa_bo->offset = offset; sa_bo->size = size; list_add(&sa_bo->list, head); + mutex_unlock(&sa_manager->mutex); return 0; }

void radeon_sa_bo_free(struct radeon_device *rdev, struct radeon_sa_bo *sa_bo) { + mutex_lock(&sa_bo->manager->mutex); list_del_init(&sa_bo->list); + mutex_unlock(&sa_bo->manager->mutex); }

From: Christian König deathsimple@vodafone.de

Dumping the current allocations.

v2: convert to mutex

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon_object.h | 5 +++++ drivers/gpu/drm/radeon/radeon_ring.c | 22 ++++++++++++++++++++++ drivers/gpu/drm/radeon/radeon_sa.c | 14 ++++++++++++++ 3 files changed, 41 insertions(+), 0 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h index f9104be..d9b9333 100644 --- a/drivers/gpu/drm/radeon/radeon_object.h +++ b/drivers/gpu/drm/radeon/radeon_object.h @@ -161,5 +161,10 @@ extern int radeon_sa_bo_new(struct radeon_device *rdev, unsigned size, unsigned align); extern void radeon_sa_bo_free(struct radeon_device *rdev, struct radeon_sa_bo *sa_bo); +#if defined(CONFIG_DEBUG_FS) +extern void radeon_sa_bo_dump_debug_info(struct radeon_sa_manager *sa_manager, + struct seq_file *m); +#endif +

#endif diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 1b020ef..1d9bce9 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -529,6 +529,23 @@ static int radeon_debugfs_ib_info(struct seq_file *m, void *data) static struct drm_info_list radeon_debugfs_ib_list[RADEON_IB_POOL_SIZE]; static char radeon_debugfs_ib_names[RADEON_IB_POOL_SIZE][32]; static unsigned radeon_debugfs_ib_idx[RADEON_IB_POOL_SIZE]; + +static int radeon_debugfs_sa_info(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *) m->private; + struct drm_device *dev = node->minor->dev; + struct radeon_device *rdev = dev->dev_private; + + radeon_sa_bo_dump_debug_info(&rdev->ib_pool.sa_manager, m); + + return 0; + +} + +static struct drm_info_list radeon_debugfs_sa_list[] = { + {"radeon_sa_info", &radeon_debugfs_sa_info, 0, NULL}, +}; + #endif

int radeon_debugfs_ring_init(struct radeon_device *rdev, struct radeon_ring *ring) @@ -555,6 +572,11 @@ int radeon_debugfs_ib_init(struct radeon_device *rdev) { #if defined(CONFIG_DEBUG_FS) unsigned i; + int r; + + r = radeon_debugfs_add_files(rdev, radeon_debugfs_sa_list, 1); + if (r) + return r;

for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { sprintf(radeon_debugfs_ib_names[i], "radeon_ib_%04u", i); diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c index 8c76c2e..88906c9 100644 --- a/drivers/gpu/drm/radeon/radeon_sa.c +++ b/drivers/gpu/drm/radeon/radeon_sa.c @@ -190,3 +190,17 @@ void radeon_sa_bo_free(struct radeon_device *rdev, struct radeon_sa_bo *sa_bo) list_del_init(&sa_bo->list); mutex_unlock(&sa_bo->manager->mutex); } + +#if defined(CONFIG_DEBUG_FS) +void radeon_sa_bo_dump_debug_info(struct radeon_sa_manager *sa_manager, + struct seq_file *m) +{ + struct radeon_sa_bo *i; + + mutex_lock(&sa_manager->mutex); + list_for_each_entry(i, &sa_manager->sa_bo, list) { + seq_printf(m, "offset %08d: size %4d\n", i->offset, i->size); + } + mutex_unlock(&sa_manager->mutex); +} +#endif

From: Jerome Glisse jglisse@redhat.com

wakequeue is use in case we want to wait until we got something that allow to allocate the object.

Signed-off-by: Christian König deathsimple@vodafone.de Signed-off-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 1 + drivers/gpu/drm/radeon/radeon_gart.c | 2 +- drivers/gpu/drm/radeon/radeon_object.h | 2 +- drivers/gpu/drm/radeon/radeon_ring.c | 6 +++++- drivers/gpu/drm/radeon/radeon_sa.c | 32 ++++++++++++++++++++++++++++++-- 5 files changed, 38 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index ad12ef8..8a6dd46 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -382,6 +382,7 @@ struct radeon_bo_list { */ struct radeon_sa_manager { struct mutex mutex; + wait_queue_head_t queue; struct radeon_bo *bo; struct list_head sa_bo; unsigned size; diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c index c58a036..7af4ff9 100644 --- a/drivers/gpu/drm/radeon/radeon_gart.c +++ b/drivers/gpu/drm/radeon/radeon_gart.c @@ -395,7 +395,7 @@ int radeon_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm) retry: r = radeon_sa_bo_new(rdev, &rdev->vm_manager.sa_manager, &vm->sa_bo, RADEON_GPU_PAGE_ALIGN(vm->last_pfn * 8), - RADEON_GPU_PAGE_SIZE); + RADEON_GPU_PAGE_SIZE, false); if (r) { if (list_empty(&rdev->vm_manager.lru_vm)) { return r; diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h index d9b9333..85f33d9 100644 --- a/drivers/gpu/drm/radeon/radeon_object.h +++ b/drivers/gpu/drm/radeon/radeon_object.h @@ -158,7 +158,7 @@ extern int radeon_sa_bo_manager_suspend(struct radeon_device *rdev, extern int radeon_sa_bo_new(struct radeon_device *rdev, struct radeon_sa_manager *sa_manager, struct radeon_sa_bo *sa_bo, - unsigned size, unsigned align); + unsigned size, unsigned align, bool block); extern void radeon_sa_bo_free(struct radeon_device *rdev, struct radeon_sa_bo *sa_bo); #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 1d9bce9..f8ecb3e 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -124,7 +124,7 @@ retry: if (rdev->ib_pool.ibs[idx].fence == NULL) { r = radeon_sa_bo_new(rdev, &rdev->ib_pool.sa_manager, &rdev->ib_pool.ibs[idx].sa_bo, - size, 256); + size, 256, false); if (!r) { *ib = &rdev->ib_pool.ibs[idx]; (*ib)->ptr = rdev->ib_pool.sa_manager.cpu_ptr; @@ -222,6 +222,10 @@ int radeon_ib_pool_init(struct radeon_device *rdev) return 0; }

+ /* copy over the sa_manager we have to do this to silence + * kernel possible deadlock detection code to complain about + * a false positive + */ rdev->ib_pool.sa_manager = tmp; INIT_LIST_HEAD(&rdev->ib_pool.sa_manager.sa_bo); for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c index 8c0b3e6..f7b20b1 100644 --- a/drivers/gpu/drm/radeon/radeon_sa.c +++ b/drivers/gpu/drm/radeon/radeon_sa.c @@ -38,6 +38,7 @@ int radeon_sa_bo_manager_init(struct radeon_device *rdev, int r;

mutex_init(&sa_manager->mutex); + init_waitqueue_head(&sa_manager->queue); sa_manager->bo = NULL; sa_manager->size = size; sa_manager->shole_size = 0; @@ -61,6 +62,8 @@ void radeon_sa_bo_manager_fini(struct radeon_device *rdev, { struct radeon_sa_bo *sa_bo, *tmp;

+ wake_up_all(&sa_manager->queue); + mutex_lock(&sa_manager->mutex); if (!list_empty(&sa_manager->sa_bo)) { dev_err(rdev->dev, "sa_manager is not empty, clearing anyway\n"); } @@ -69,6 +72,7 @@ void radeon_sa_bo_manager_fini(struct radeon_device *rdev, } radeon_bo_unref(&sa_manager->bo); sa_manager->size = 0; + mutex_unlock(&sa_manager->mutex); }

int radeon_sa_bo_manager_start(struct radeon_device *rdev, @@ -117,6 +121,17 @@ int radeon_sa_bo_manager_suspend(struct radeon_device *rdev, return r; }

+static inline unsigned radeon_sa_bo_max_free(struct radeon_sa_manager *m, + unsigned align) +{ + unsigned start, end, wasted; + + wasted = (align - (sa_manager->ehole_offset % align)) % align; + end = wasted > sa_manager->ehole_size ? 0 : sa_manager->ehole_size - wasted; + start = sa_manager->shole_size; + return start < end ? end : start; +} + /* * Principe is simple, we keep a list of sub allocation in offset * order (first entry has offset == 0, last entry has the highest @@ -135,13 +150,14 @@ int radeon_sa_bo_manager_suspend(struct radeon_device *rdev, int radeon_sa_bo_new(struct radeon_device *rdev, struct radeon_sa_manager *sa_manager, struct radeon_sa_bo *sa_bo, - unsigned size, unsigned align) + unsigned size, unsigned align, bool block) { unsigned offset = 0, wasted = 0;

BUG_ON(align > RADEON_GPU_PAGE_SIZE); BUG_ON(size > sa_manager->size);

+retry: mutex_lock(&sa_manager->mutex);

wasted = (align - (sa_manager->ehole_offset % align)) % align; @@ -161,8 +177,20 @@ int radeon_sa_bo_new(struct radeon_device *rdev, offset = 0; goto out; } - mutex_unlock(&sa_manager->mutex); + + if (block) { + /* failed to find something big enough, wait + * for the biggest hole to increase in size + */ + r = wait_event_interruptible(sa_manager->queue, + radeon_sa_bo_max_free(sa_manager, align) >= size); + if (r) { + return r; + } + goto retry; + } + return -ENOMEM;

out:

From: Christian König deathsimple@vodafone.de

Directly use the suballocator to get small chunks of memory. It's equally fast and doesn't crash when we encounter a GPU reset.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/evergreen.c | 1 - drivers/gpu/drm/radeon/ni.c | 1 - drivers/gpu/drm/radeon/r600.c | 1 - drivers/gpu/drm/radeon/radeon.h | 29 +------ drivers/gpu/drm/radeon/radeon_device.c | 2 - drivers/gpu/drm/radeon/radeon_semaphore.c | 134 ++++------------------------- drivers/gpu/drm/radeon/rv770.c | 1 - drivers/gpu/drm/radeon/si.c | 1 - 8 files changed, 22 insertions(+), 148 deletions(-)

diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index ca47f52..a76389c 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -3431,7 +3431,6 @@ void evergreen_fini(struct radeon_device *rdev) evergreen_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); - radeon_semaphore_driver_fini(rdev); radeon_fence_driver_fini(rdev); radeon_agp_fini(rdev); radeon_bo_fini(rdev); diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 0146428..c0b0956 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1773,7 +1773,6 @@ void cayman_fini(struct radeon_device *rdev) cayman_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); - radeon_semaphore_driver_fini(rdev); radeon_fence_driver_fini(rdev); radeon_bo_fini(rdev); radeon_atombios_fini(rdev); diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 06eccd1..a252160 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2672,7 +2672,6 @@ void r600_fini(struct radeon_device *rdev) r600_vram_scratch_fini(rdev); radeon_agp_fini(rdev); radeon_gem_fini(rdev); - radeon_semaphore_driver_fini(rdev); radeon_fence_driver_fini(rdev); radeon_bo_fini(rdev); radeon_atombios_fini(rdev); diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 8a6dd46..83ff115 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -432,34 +432,12 @@ int radeon_mode_dumb_destroy(struct drm_file *file_priv, /* * Semaphores. */ -struct radeon_ring; - -#define RADEON_SEMAPHORE_BO_SIZE 256 - -struct radeon_semaphore_driver { - rwlock_t lock; - struct list_head bo; -}; - -struct radeon_semaphore_bo; - -/* everything here is constant */ struct radeon_semaphore { - struct list_head list; - uint64_t gpu_addr; - uint32_t *cpu_ptr; - struct radeon_semaphore_bo *bo; -}; - -struct radeon_semaphore_bo { - struct list_head list; - struct radeon_ib *ib; - struct list_head free; - struct radeon_semaphore semaphores[RADEON_SEMAPHORE_BO_SIZE/8]; - unsigned nused; + struct radeon_sa_bo sa_bo; + signed waiters; + uint64_t gpu_addr; };

-void radeon_semaphore_driver_fini(struct radeon_device *rdev); int radeon_semaphore_create(struct radeon_device *rdev, struct radeon_semaphore **semaphore); void radeon_semaphore_emit_signal(struct radeon_device *rdev, int ring, @@ -1523,7 +1501,6 @@ struct radeon_device { struct radeon_mman mman; rwlock_t fence_lock; struct radeon_fence_driver fence_drv[RADEON_NUM_RINGS]; - struct radeon_semaphore_driver semaphore_drv; struct radeon_ring ring[RADEON_NUM_RINGS]; struct radeon_ib_pool ib_pool; struct radeon_irq irq; diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index eb63a06..f314819 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -733,11 +733,9 @@ int radeon_device_init(struct radeon_device *rdev, mutex_init(&rdev->pm.mutex); mutex_init(&rdev->vram_mutex); rwlock_init(&rdev->fence_lock); - rwlock_init(&rdev->semaphore_drv.lock); INIT_LIST_HEAD(&rdev->gem.objects); init_waitqueue_head(&rdev->irq.vblank_queue); init_waitqueue_head(&rdev->irq.idle_queue); - INIT_LIST_HEAD(&rdev->semaphore_drv.bo); /* initialize vm here */ rdev->vm_manager.use_bitmap = 1; rdev->vm_manager.max_pfn = 1 << 20; diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/radeon/radeon_semaphore.c index 61dd4e3..a746410 100644 --- a/drivers/gpu/drm/radeon/radeon_semaphore.c +++ b/drivers/gpu/drm/radeon/radeon_semaphore.c @@ -31,108 +31,27 @@ #include "drm.h" #include "radeon.h"

-static int radeon_semaphore_add_bo(struct radeon_device *rdev) +int radeon_semaphore_create(struct radeon_device *rdev, + struct radeon_semaphore **semaphore) { - struct radeon_semaphore_bo *bo; - unsigned long irq_flags; - uint64_t gpu_addr; - uint32_t *cpu_ptr; - int r, i; - + int r;

- bo = kmalloc(sizeof(struct radeon_semaphore_bo), GFP_KERNEL); - if (bo == NULL) { + *semaphore = kmalloc(sizeof(struct radeon_semaphore), GFP_KERNEL); + if (*semaphore == NULL) { return -ENOMEM; } - INIT_LIST_HEAD(&bo->free); - INIT_LIST_HEAD(&bo->list); - bo->nused = 0;

- r = radeon_ib_get(rdev, 0, &bo->ib, RADEON_SEMAPHORE_BO_SIZE); + r = radeon_sa_bo_new(rdev, &rdev->ib_pool.sa_manager, + &(*semaphore)->sa_bo, 8, 8, true); if (r) { - dev_err(rdev->dev, "failed to get a bo after 5 retry\n"); - kfree(bo); + kfree(*semaphore); + *semaphore = NULL; return r; } - gpu_addr = rdev->ib_pool.sa_manager.gpu_addr; - gpu_addr += bo->ib->sa_bo.offset; - cpu_ptr = rdev->ib_pool.sa_manager.cpu_ptr; - cpu_ptr += (bo->ib->sa_bo.offset >> 2); - for (i = 0; i < (RADEON_SEMAPHORE_BO_SIZE/8); i++) { - bo->semaphores[i].gpu_addr = gpu_addr; - bo->semaphores[i].cpu_ptr = cpu_ptr; - bo->semaphores[i].bo = bo; - list_add_tail(&bo->semaphores[i].list, &bo->free); - gpu_addr += 8; - cpu_ptr += 2; - } - write_lock_irqsave(&rdev->semaphore_drv.lock, irq_flags); - list_add_tail(&bo->list, &rdev->semaphore_drv.bo); - write_unlock_irqrestore(&rdev->semaphore_drv.lock, irq_flags); - return 0; -}

-static void radeon_semaphore_del_bo_locked(struct radeon_device *rdev, - struct radeon_semaphore_bo *bo) -{ - radeon_sa_bo_free(rdev, &bo->ib->sa_bo); - radeon_fence_unref(&bo->ib->fence); - list_del(&bo->list); - kfree(bo); -} - -void radeon_semaphore_shrink_locked(struct radeon_device *rdev) -{ - struct radeon_semaphore_bo *bo, *n; - - if (list_empty(&rdev->semaphore_drv.bo)) { - return; - } - /* only shrink if first bo has free semaphore */ - bo = list_first_entry(&rdev->semaphore_drv.bo, struct radeon_semaphore_bo, list); - if (list_empty(&bo->free)) { - return; - } - list_for_each_entry_safe_continue(bo, n, &rdev->semaphore_drv.bo, list) { - if (bo->nused) - continue; - radeon_semaphore_del_bo_locked(rdev, bo); - } -} - -int radeon_semaphore_create(struct radeon_device *rdev, - struct radeon_semaphore **semaphore) -{ - struct radeon_semaphore_bo *bo; - unsigned long irq_flags; - bool do_retry = true; - int r; - -retry: - *semaphore = NULL; - write_lock_irqsave(&rdev->semaphore_drv.lock, irq_flags); - list_for_each_entry(bo, &rdev->semaphore_drv.bo, list) { - if (list_empty(&bo->free)) - continue; - *semaphore = list_first_entry(&bo->free, struct radeon_semaphore, list); - (*semaphore)->cpu_ptr[0] = 0; - (*semaphore)->cpu_ptr[1] = 0; - list_del(&(*semaphore)->list); - bo->nused++; - break; - } - write_unlock_irqrestore(&rdev->semaphore_drv.lock, irq_flags); - - if (*semaphore == NULL) { - if (do_retry) { - do_retry = false; - r = radeon_semaphore_add_bo(rdev); - if (r) - return r; - goto retry; - } - return -ENOMEM; - } + (*semaphore)->waiters = 0; + (*semaphore)->gpu_addr = radeon_sa_bo_gpu_addr(&(*semaphore)->sa_bo); + *((uint64_t*)radeon_sa_bo_cpu_addr(&(*semaphore)->sa_bo)) = 0;

return 0; } @@ -140,39 +59,24 @@ retry: void radeon_semaphore_emit_signal(struct radeon_device *rdev, int ring, struct radeon_semaphore *semaphore) { + --semaphore->waiters; radeon_semaphore_ring_emit(rdev, ring, &rdev->ring[ring], semaphore, false); }

void radeon_semaphore_emit_wait(struct radeon_device *rdev, int ring, struct radeon_semaphore *semaphore) { + ++semaphore->waiters; radeon_semaphore_ring_emit(rdev, ring, &rdev->ring[ring], semaphore, true); }

void radeon_semaphore_free(struct radeon_device *rdev, struct radeon_semaphore *semaphore) { - unsigned long irq_flags; - - write_lock_irqsave(&rdev->semaphore_drv.lock, irq_flags); - semaphore->bo->nused--; - list_add_tail(&semaphore->list, &semaphore->bo->free); - radeon_semaphore_shrink_locked(rdev); - write_unlock_irqrestore(&rdev->semaphore_drv.lock, irq_flags); -} - -void radeon_semaphore_driver_fini(struct radeon_device *rdev) -{ - struct radeon_semaphore_bo *bo, *n; - unsigned long irq_flags; - - write_lock_irqsave(&rdev->semaphore_drv.lock, irq_flags); - /* we force to free everything */ - list_for_each_entry_safe(bo, n, &rdev->semaphore_drv.bo, list) { - if (!list_empty(&bo->free)) { - dev_err(rdev->dev, "still in use semaphore\n"); - } - radeon_semaphore_del_bo_locked(rdev, bo); + if (semaphore->waiters > 0) { + dev_err(rdev->dev, "Semaphore %p has more waiters than signalers," + " hardware lockup imminent!\n", semaphore); } - write_unlock_irqrestore(&rdev->semaphore_drv.lock, irq_flags); + radeon_sa_bo_free(rdev, &semaphore->sa_bo); + kfree(semaphore); } diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index a8b0016..40f82e2 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1278,7 +1278,6 @@ void rv770_fini(struct radeon_device *rdev) rv770_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); - radeon_semaphore_driver_fini(rdev); radeon_fence_driver_fini(rdev); radeon_agp_fini(rdev); radeon_bo_fini(rdev); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 27bda98..f05d37a 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -4117,7 +4117,6 @@ void si_fini(struct radeon_device *rdev) si_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); - radeon_semaphore_driver_fini(rdev); radeon_fence_driver_fini(rdev); radeon_bo_fini(rdev); radeon_atombios_fini(rdev);

From: Christian König deathsimple@vodafone.de

As discussed with Michel that name better describes the behavior of this function.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Michel Dänzer michel.daenzer@amd.com Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 2 +- drivers/gpu/drm/radeon/radeon_device.c | 2 +- drivers/gpu/drm/radeon/radeon_fence.c | 4 ++-- 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 83ff115..fbbbc5a 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -285,7 +285,7 @@ void radeon_fence_process(struct radeon_device *rdev, int ring); bool radeon_fence_signaled(struct radeon_fence *fence); int radeon_fence_wait(struct radeon_fence *fence, bool interruptible); int radeon_fence_wait_next(struct radeon_device *rdev, int ring); -int radeon_fence_wait_last(struct radeon_device *rdev, int ring); +int radeon_fence_wait_empty(struct radeon_device *rdev, int ring); struct radeon_fence *radeon_fence_ref(struct radeon_fence *fence); void radeon_fence_unref(struct radeon_fence **fence); int radeon_fence_count_emitted(struct radeon_device *rdev, int ring); diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index f314819..d77f19f 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -913,7 +913,7 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state) radeon_bo_evict_vram(rdev); /* wait for gpu to finish processing current batch */ for (i = 0; i < RADEON_NUM_RINGS; i++) - radeon_fence_wait_last(rdev, i); + radeon_fence_wait_empty(rdev, i);

radeon_save_bios_scratch_regs(rdev);

diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 2fbbc34..2d13843 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -297,7 +297,7 @@ int radeon_fence_wait_next(struct radeon_device *rdev, int ring) return r; }

-int radeon_fence_wait_last(struct radeon_device *rdev, int ring) +int radeon_fence_wait_empty(struct radeon_device *rdev, int ring) { unsigned long irq_flags; struct radeon_fence *fence; @@ -442,7 +442,7 @@ void radeon_fence_driver_fini(struct radeon_device *rdev) for (ring = 0; ring < RADEON_NUM_RINGS; ring++) { if (!rdev->fence_drv[ring].initialized) continue; - radeon_fence_wait_last(rdev, ring); + radeon_fence_wait_empty(rdev, ring); wake_up_all(&rdev->fence_drv[ring].queue); write_lock_irqsave(&rdev->fence_lock, irq_flags); radeon_scratch_free(rdev, rdev->fence_drv[ring].scratch_reg);

From: Christian König deathsimple@vodafone.de

It's never used and so practically superfluous.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 1 - drivers/gpu/drm/radeon/radeon_fence.c | 7 ------- 2 files changed, 0 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 11cceb3..fa4dc73 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -257,7 +257,6 @@ struct radeon_fence_driver { uint32_t last_seq; unsigned long last_activity; wait_queue_head_t queue; - struct list_head created; struct list_head emitted; bool initialized; }; diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index c58660a..326bdb5 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -154,8 +154,6 @@ int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence **fence, int ring) { - unsigned long irq_flags; - *fence = kmalloc(sizeof(struct radeon_fence), GFP_KERNEL); if ((*fence) == NULL) { return -ENOMEM; @@ -170,10 +168,6 @@ int radeon_fence_create(struct radeon_device *rdev, (*fence)->signal_callback = NULL; (*fence)->callback_data = NULL; INIT_LIST_HEAD(&(*fence)->list); - - write_lock_irqsave(&rdev->fence_lock, irq_flags); - list_add_tail(&(*fence)->list, &rdev->fence_drv[ring].created); - write_unlock_irqrestore(&rdev->fence_lock, irq_flags); return 0; }

@@ -452,7 +446,6 @@ static void radeon_fence_driver_init_ring(struct radeon_device *rdev, int ring) rdev->fence_drv[ring].cpu_addr = NULL; rdev->fence_drv[ring].gpu_addr = 0; atomic_set(&rdev->fence_drv[ring].seq, 0); - INIT_LIST_HEAD(&rdev->fence_drv[ring].created); INIT_LIST_HEAD(&rdev->fence_drv[ring].emitted); init_waitqueue_head(&rdev->fence_drv[ring].queue); rdev->fence_drv[ring].initialized = false;

From: Christian König deathsimple@vodafone.de

It isn't necessary any more and the suballocator seems to perform even better.

v2: ignore ERESTARTSYS in error reporting, split fence changes into seperate patch, use try_free SA callback to avoid lockups v3: rebase on top of sa manager new patch

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon.h | 17 +-- drivers/gpu/drm/radeon/radeon_device.c | 1 - drivers/gpu/drm/radeon/radeon_gart.c | 12 +- drivers/gpu/drm/radeon/radeon_ring.c | 274 ++++++++++------------------- drivers/gpu/drm/radeon/radeon_semaphore.c | 2 +- 5 files changed, 103 insertions(+), 203 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index f18aa13..0c8893d 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -614,7 +614,6 @@ void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc);

struct radeon_ib { struct radeon_sa_bo sa_bo; - unsigned idx; uint32_t length_dw; uint64_t gpu_addr; uint32_t *ptr; @@ -623,18 +622,6 @@ struct radeon_ib { bool is_const_ib; };

-/* - * locking - - * mutex protects scheduled_ibs, ready, alloc_bm - */ -struct radeon_ib_pool { - struct radeon_mutex mutex; - struct radeon_sa_manager sa_manager; - struct radeon_ib ibs[RADEON_IB_POOL_SIZE]; - bool ready; - unsigned head_id; -}; - struct radeon_ring { struct radeon_bo *ring_obj; volatile uint32_t *ring; @@ -775,7 +762,6 @@ struct si_rlc { int radeon_ib_get(struct radeon_device *rdev, int ring, struct radeon_ib **ib, unsigned size); void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib); -bool radeon_ib_try_free(struct radeon_device *rdev, struct radeon_ib *ib); int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib); int radeon_ib_pool_init(struct radeon_device *rdev); void radeon_ib_pool_fini(struct radeon_device *rdev); @@ -1508,7 +1494,8 @@ struct radeon_device { rwlock_t fence_lock; struct radeon_fence_driver fence_drv[RADEON_NUM_RINGS]; struct radeon_ring ring[RADEON_NUM_RINGS]; - struct radeon_ib_pool ib_pool; + bool ib_pool_ready; + struct radeon_sa_manager sa_manager; struct radeon_irq irq; struct radeon_asic *asic; struct radeon_gem gem; diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index d77f19f..26fb9da 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -723,7 +723,6 @@ int radeon_device_init(struct radeon_device *rdev, /* mutex initialization are all done here so we * can recall function without having locking issues */ radeon_mutex_init(&rdev->cs_mutex); - radeon_mutex_init(&rdev->ib_pool.mutex); for (i = 0; i < RADEON_NUM_RINGS; ++i) mutex_init(&rdev->ring[i].mutex); mutex_init(&rdev->dc_hw_i2c_mutex); diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c index 1766671..8eba1c6 100644 --- a/drivers/gpu/drm/radeon/radeon_gart.c +++ b/drivers/gpu/drm/radeon/radeon_gart.c @@ -432,8 +432,8 @@ retry_id: rdev->vm_manager.use_bitmap |= 1 << id; vm->id = id; list_add_tail(&vm->list, &rdev->vm_manager.lru_vm); - return radeon_vm_bo_update_pte(rdev, vm, rdev->ib_pool.sa_manager.bo, - &rdev->ib_pool.sa_manager.bo->tbo.mem); + return radeon_vm_bo_update_pte(rdev, vm, rdev->sa_manager.bo, + &rdev->sa_manager.bo->tbo.mem); }

/* object have to be reserved */ @@ -631,7 +631,7 @@ int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm) /* map the ib pool buffer at 0 in virtual address space, set * read only */ - r = radeon_vm_bo_add(rdev, vm, rdev->ib_pool.sa_manager.bo, 0, + r = radeon_vm_bo_add(rdev, vm, rdev->sa_manager.bo, 0, RADEON_VM_PAGE_READABLE | RADEON_VM_PAGE_SNOOPED); return r; } @@ -648,12 +648,12 @@ void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm) radeon_mutex_unlock(&rdev->cs_mutex);

/* remove all bo */ - r = radeon_bo_reserve(rdev->ib_pool.sa_manager.bo, false); + r = radeon_bo_reserve(rdev->sa_manager.bo, false); if (!r) { - bo_va = radeon_bo_va(rdev->ib_pool.sa_manager.bo, vm); + bo_va = radeon_bo_va(rdev->sa_manager.bo, vm); list_del_init(&bo_va->bo_list); list_del_init(&bo_va->vm_list); - radeon_bo_unreserve(rdev->ib_pool.sa_manager.bo); + radeon_bo_unreserve(rdev->sa_manager.bo); kfree(bo_va); } if (!list_empty(&vm->va)) { diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 0293342..7031d69 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -24,6 +24,7 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse + * Christian König */ #include <linux/seq_file.h> #include <linux/slab.h> @@ -33,8 +34,10 @@ #include "radeon.h" #include "atom.h"

-int radeon_debugfs_ib_init(struct radeon_device *rdev); -int radeon_debugfs_ring_init(struct radeon_device *rdev, struct radeon_ring *ring); +/* + * IB. + */ +int radeon_debugfs_sa_init(struct radeon_device *rdev);

u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx) { @@ -61,121 +64,71 @@ u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx) return idx_value; }

-void radeon_ring_write(struct radeon_ring *ring, uint32_t v) -{ -#if DRM_DEBUG_CODE - if (ring->count_dw <= 0) { - DRM_ERROR("radeon: writting more dword to ring than expected !\n"); - } -#endif - ring->ring[ring->wptr++] = v; - ring->wptr &= ring->ptr_mask; - ring->count_dw--; - ring->ring_free_dw--; -} - -/* - * IB. - */ -bool radeon_ib_try_free(struct radeon_device *rdev, struct radeon_ib *ib) -{ - bool done = false; - - /* only free ib which have been emited */ - if (ib->fence && ib->fence->emitted) { - if (radeon_fence_signaled(ib->fence)) { - radeon_fence_unref(&ib->fence); - radeon_sa_bo_free(rdev, &ib->sa_bo); - done = true; - } - } - return done; -} - int radeon_ib_get(struct radeon_device *rdev, int ring, struct radeon_ib **ib, unsigned size) { - struct radeon_fence *fence; - unsigned cretry = 0; - int r = 0, i, idx; + int r;

- *ib = NULL; - /* align size on 256 bytes */ - size = ALIGN(size, 256); + *ib = kmalloc(sizeof(struct radeon_ib), GFP_KERNEL); + if (*ib == NULL) { + return -ENOMEM; + }

- r = radeon_fence_create(rdev, &fence, ring); + r = radeon_sa_bo_new(rdev, &rdev->sa_manager, &(*ib)->sa_bo, size, 256, true); if (r) { - dev_err(rdev->dev, "failed to create fence for new IB\n"); + if (r != -ERESTARTSYS) { + dev_err(rdev->dev, "failed to get a new IB (%d)\n", r); + } + kfree(*ib); + *ib = NULL; return r; }

- radeon_mutex_lock(&rdev->ib_pool.mutex); - idx = rdev->ib_pool.head_id; -retry: - if (cretry > 5) { - dev_err(rdev->dev, "failed to get an ib after 5 retry\n"); - radeon_mutex_unlock(&rdev->ib_pool.mutex); - radeon_fence_unref(&fence); - return -ENOMEM; - } - cretry++; - for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { - radeon_ib_try_free(rdev, &rdev->ib_pool.ibs[idx]); - if (rdev->ib_pool.ibs[idx].fence == NULL) { - r = radeon_sa_bo_new(rdev, &rdev->ib_pool.sa_manager, - &rdev->ib_pool.ibs[idx].sa_bo, - size, 256, false); - if (!r) { - *ib = &rdev->ib_pool.ibs[idx]; - (*ib)->ptr = radeon_sa_bo_cpu_addr(&(*ib)->sa_bo); - (*ib)->gpu_addr = radeon_sa_bo_gpu_addr(&(*ib)->sa_bo); - (*ib)->fence = fence; - (*ib)->vm_id = 0; - (*ib)->is_const_ib = false; - /* ib are most likely to be allocated in a ring fashion - * thus rdev->ib_pool.head_id should be the id of the - * oldest ib - */ - rdev->ib_pool.head_id = (1 + idx); - rdev->ib_pool.head_id &= (RADEON_IB_POOL_SIZE - 1); - radeon_mutex_unlock(&rdev->ib_pool.mutex); - return 0; - } - } - idx = (idx + 1) & (RADEON_IB_POOL_SIZE - 1); - } - /* this should be rare event, ie all ib scheduled none signaled yet. - */ - for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { - if (rdev->ib_pool.ibs[idx].fence && rdev->ib_pool.ibs[idx].fence->emitted) { - r = radeon_fence_wait(rdev->ib_pool.ibs[idx].fence, false); - if (!r) { - goto retry; - } - /* an error happened */ - break; - } - idx = (idx + 1) & (RADEON_IB_POOL_SIZE - 1); + r = radeon_fence_create(rdev, &(*ib)->fence, ring); + if (r) { + dev_err(rdev->dev, "failed to create fence for new IB (%d)\n", r); + radeon_sa_bo_free(rdev, &(*ib)->sa_bo); + kfree(*ib); + *ib = NULL; + return r; } - radeon_mutex_unlock(&rdev->ib_pool.mutex); - radeon_fence_unref(&fence); - return r; + + (*ib)->ptr = rdev->sa_manager.cpu_ptr; + (*ib)->ptr += ((*ib)->sa_bo.offset >> 2); + (*ib)->gpu_addr = rdev->sa_manager.gpu_addr; + (*ib)->gpu_addr += (*ib)->sa_bo.offset; + (*ib)->vm_id = 0; + (*ib)->is_const_ib = false; + + return 0; +} + +static void radeon_ib_destroy(struct radeon_device *rdev, void *data) +{ + struct radeon_ib *ib = data; + radeon_sa_bo_free(rdev, &ib->sa_bo); + radeon_fence_unref(&ib->fence); + kfree(ib); }

void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib) { struct radeon_ib *tmp = *ib; + bool destroy = true;

*ib = NULL; if (tmp == NULL) { return; } - radeon_mutex_lock(&rdev->ib_pool.mutex); - if (tmp->fence && !tmp->fence->emitted) { - radeon_sa_bo_free(rdev, &tmp->sa_bo); - radeon_fence_unref(&tmp->fence); + + if (tmp->fence) { + destroy = !radeon_fence_set_signal_callback( + tmp->fence, radeon_ib_destroy, tmp); + } + + if (destroy) { + radeon_ib_destroy(rdev, tmp); } - radeon_mutex_unlock(&rdev->ib_pool.mutex); }

int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib) @@ -185,7 +138,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib)

if (!ib->length_dw || !ring->ready) { /* TODO: Nothings in the ib we should report. */ - DRM_ERROR("radeon: couldn't schedule IB(%u).\n", ib->idx); + DRM_ERROR("radeon: couldn't schedule IB.\n"); return -EINVAL; }

@@ -201,72 +154,58 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib) return 0; }

+static void radeon_ib_try_free(struct radeon_device *rdev) +{ + int ring; + + for (ring = 0; ring < RADEON_NUM_RINGS; ring++) { + if (rdev->fence_drv[ring].initialized) + radeon_fence_process(rdev, ring); + } +} + int radeon_ib_pool_init(struct radeon_device *rdev) { - struct radeon_sa_manager tmp; - int i, r; + int r;

- r = radeon_sa_bo_manager_init(rdev, &tmp, + if (rdev->ib_pool_ready) { + return 0; + } + + r = radeon_sa_bo_manager_init(rdev, &rdev->sa_manager, RADEON_IB_POOL_SIZE*64*1024, - RADEON_GEM_DOMAIN_GTT, NULL); + RADEON_GEM_DOMAIN_GTT, + radeon_ib_try_free); if (r) { return r; }

- radeon_mutex_lock(&rdev->ib_pool.mutex); - if (rdev->ib_pool.ready) { - radeon_mutex_unlock(&rdev->ib_pool.mutex); - radeon_sa_bo_manager_fini(rdev, &tmp); - return 0; + if (radeon_debugfs_sa_init(rdev)) { + DRM_ERROR("Failed to register debugfs file for SA !\n"); }

- /* copy over the sa_manager we have to do this to silence - * kernel possible deadlock detection code to complain about - * a false positive - */ - rdev->ib_pool.sa_manager = tmp; - INIT_LIST_HEAD(&rdev->ib_pool.sa_manager.sa_bo); - for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { - rdev->ib_pool.ibs[i].fence = NULL; - rdev->ib_pool.ibs[i].idx = i; - rdev->ib_pool.ibs[i].length_dw = 0; - INIT_LIST_HEAD(&rdev->ib_pool.ibs[i].sa_bo.list); - } - rdev->ib_pool.head_id = 0; - rdev->ib_pool.ready = true; DRM_INFO("radeon: ib pool ready.\n"); + rdev->ib_pool_ready = true;

- if (radeon_debugfs_ib_init(rdev)) { - DRM_ERROR("Failed to register debugfs file for IB !\n"); - } - radeon_mutex_unlock(&rdev->ib_pool.mutex); return 0; }

void radeon_ib_pool_fini(struct radeon_device *rdev) { - unsigned i; - - radeon_mutex_lock(&rdev->ib_pool.mutex); - if (rdev->ib_pool.ready) { - for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { - radeon_sa_bo_free(rdev, &rdev->ib_pool.ibs[i].sa_bo); - radeon_fence_unref(&rdev->ib_pool.ibs[i].fence); - } - radeon_sa_bo_manager_fini(rdev, &rdev->ib_pool.sa_manager); - rdev->ib_pool.ready = false; + if (rdev->ib_pool_ready) { + radeon_sa_bo_manager_fini(rdev, &rdev->sa_manager); + rdev->ib_pool_ready = false; } - radeon_mutex_unlock(&rdev->ib_pool.mutex); }

int radeon_ib_pool_start(struct radeon_device *rdev) { - return radeon_sa_bo_manager_start(rdev, &rdev->ib_pool.sa_manager); + return radeon_sa_bo_manager_start(rdev, &rdev->sa_manager); }

int radeon_ib_pool_suspend(struct radeon_device *rdev) { - return radeon_sa_bo_manager_suspend(rdev, &rdev->ib_pool.sa_manager); + return radeon_sa_bo_manager_suspend(rdev, &rdev->sa_manager); }

int radeon_ib_ring_tests(struct radeon_device *rdev) @@ -302,6 +241,21 @@ int radeon_ib_ring_tests(struct radeon_device *rdev) /* * Ring. */ +int radeon_debugfs_ring_init(struct radeon_device *rdev, struct radeon_ring *ring); + +void radeon_ring_write(struct radeon_ring *ring, uint32_t v) +{ +#if DRM_DEBUG_CODE + if (ring->count_dw <= 0) { + DRM_ERROR("radeon: writting more dword to ring than expected !\n"); + } +#endif + ring->ring[ring->wptr++] = v; + ring->wptr &= ring->ptr_mask; + ring->count_dw--; + ring->ring_free_dw--; +} + int radeon_ring_index(struct radeon_device *rdev, struct radeon_ring *ring) { /* r1xx-r5xx only has CP ring */ @@ -508,37 +462,13 @@ static struct drm_info_list radeon_debugfs_ring_info_list[] = { {"radeon_ring_cp2", radeon_debugfs_ring_info, 0, &cayman_ring_type_cp2_index}, };

-static int radeon_debugfs_ib_info(struct seq_file *m, void *data) -{ - struct drm_info_node *node = (struct drm_info_node *) m->private; - struct drm_device *dev = node->minor->dev; - struct radeon_device *rdev = dev->dev_private; - struct radeon_ib *ib = &rdev->ib_pool.ibs[*((unsigned*)node->info_ent->data)]; - unsigned i; - - if (ib == NULL) { - return 0; - } - seq_printf(m, "IB %04u\n", ib->idx); - seq_printf(m, "IB fence %p\n", ib->fence); - seq_printf(m, "IB size %05u dwords\n", ib->length_dw); - for (i = 0; i < ib->length_dw; i++) { - seq_printf(m, "[%05u]=0x%08X\n", i, ib->ptr[i]); - } - return 0; -} - -static struct drm_info_list radeon_debugfs_ib_list[RADEON_IB_POOL_SIZE]; -static char radeon_debugfs_ib_names[RADEON_IB_POOL_SIZE][32]; -static unsigned radeon_debugfs_ib_idx[RADEON_IB_POOL_SIZE]; - static int radeon_debugfs_sa_info(struct seq_file *m, void *data) { struct drm_info_node *node = (struct drm_info_node *) m->private; struct drm_device *dev = node->minor->dev; struct radeon_device *rdev = dev->dev_private;

- radeon_sa_bo_dump_debug_info(&rdev->ib_pool.sa_manager, m); + radeon_sa_bo_dump_debug_info(&rdev->sa_manager, m);

return 0;

@@ -570,26 +500,10 @@ int radeon_debugfs_ring_init(struct radeon_device *rdev, struct radeon_ring *rin return 0; }

-int radeon_debugfs_ib_init(struct radeon_device *rdev) +int radeon_debugfs_sa_init(struct radeon_device *rdev) { #if defined(CONFIG_DEBUG_FS) - unsigned i; - int r; - - r = radeon_debugfs_add_files(rdev, radeon_debugfs_sa_list, 1); - if (r) - return r; - - for (i = 0; i < RADEON_IB_POOL_SIZE; i++) { - sprintf(radeon_debugfs_ib_names[i], "radeon_ib_%04u", i); - radeon_debugfs_ib_idx[i] = i; - radeon_debugfs_ib_list[i].name = radeon_debugfs_ib_names[i]; - radeon_debugfs_ib_list[i].show = &radeon_debugfs_ib_info; - radeon_debugfs_ib_list[i].driver_features = 0; - radeon_debugfs_ib_list[i].data = &radeon_debugfs_ib_idx[i]; - } - return radeon_debugfs_add_files(rdev, radeon_debugfs_ib_list, - RADEON_IB_POOL_SIZE); + return radeon_debugfs_add_files(rdev, radeon_debugfs_sa_list, 1); #else return 0; #endif diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/radeon/radeon_semaphore.c index a746410..4cdcef2 100644 --- a/drivers/gpu/drm/radeon/radeon_semaphore.c +++ b/drivers/gpu/drm/radeon/radeon_semaphore.c @@ -41,7 +41,7 @@ int radeon_semaphore_create(struct radeon_device *rdev, return -ENOMEM; }

- r = radeon_sa_bo_new(rdev, &rdev->ib_pool.sa_manager, + r = radeon_sa_bo_new(rdev, &rdev->sa_manager, &(*semaphore)->sa_bo, 8, 8, true); if (r) { kfree(*semaphore);

From: Christian König deathsimple@vodafone.de

Instead of all this humpy pumpy with recursive mutex (which also fixes only halve of the problem) move the actual gpu reset out of the fence code, return -EDEADLK and then reset the gpu in the calling ioctl function.

v2: Split removal of radeon_mutex into separate patch. Return -EAGAIN if reset is successful.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon_cs.c | 13 +++++++++++++ drivers/gpu/drm/radeon/radeon_device.c | 5 ----- drivers/gpu/drm/radeon/radeon_fence.c | 10 +++------- drivers/gpu/drm/radeon/radeon_gem.c | 16 ++++++++++++++++ 4 files changed, 32 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index 64b86e7..a0826bb 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -489,6 +489,16 @@ out: return r; }

+static int radeon_cs_handle_lockup(struct radeon_device *rdev, int r) +{ + if (r == -EDEADLK) { + r = radeon_gpu_reset(rdev); + if (!r) + r = -EAGAIN; + } + return r; +} + int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct radeon_device *rdev = dev->dev_private; @@ -510,6 +520,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (r) { DRM_ERROR("Failed to initialize parser !\n"); radeon_cs_parser_fini(&parser, r); + r = radeon_cs_handle_lockup(rdev, r); radeon_mutex_unlock(&rdev->cs_mutex); return r; } @@ -518,6 +529,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (r != -ERESTARTSYS) DRM_ERROR("Failed to parse relocation %d!\n", r); radeon_cs_parser_fini(&parser, r); + r = radeon_cs_handle_lockup(rdev, r); radeon_mutex_unlock(&rdev->cs_mutex); return r; } @@ -531,6 +543,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) } out: radeon_cs_parser_fini(&parser, r); + r = radeon_cs_handle_lockup(rdev, r); radeon_mutex_unlock(&rdev->cs_mutex); return r; } diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 26fb9da..1dac27d 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -984,9 +984,6 @@ int radeon_gpu_reset(struct radeon_device *rdev) int r; int resched;

- /* Prevent CS ioctl from interfering */ - radeon_mutex_lock(&rdev->cs_mutex); - radeon_save_bios_scratch_regs(rdev); /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&rdev->mman.bdev); @@ -1001,8 +998,6 @@ int radeon_gpu_reset(struct radeon_device *rdev) ttm_bo_unlock_delayed_workqueue(&rdev->mman.bdev, resched); }

- radeon_mutex_unlock(&rdev->cs_mutex); - if (r) { /* bad news, how to tell it to userspace ? */ dev_info(rdev->dev, "GPU reset failed\n"); diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 326bdb5..24a4d4b 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -264,6 +264,8 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr) /* change sequence value on all rings, so nobody else things there is a lockup */ for (i = 0; i < RADEON_NUM_RINGS; ++i) rdev->fence_drv[i].last_seq -= 0x10000; + + rdev->fence_drv[fence->ring].last_activity = jiffies; write_unlock_irqrestore(&rdev->fence_lock, irq_flags);

if (radeon_ring_is_lockup(rdev, fence->ring, &rdev->ring[fence->ring])) { @@ -274,13 +276,7 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr)

/* mark the ring as not ready any more */ rdev->ring[fence->ring].ready = false; - r = radeon_gpu_reset(rdev); - if (r) - return r; - - write_lock_irqsave(&rdev->fence_lock, irq_flags); - rdev->fence_drv[fence->ring].last_activity = jiffies; - write_unlock_irqrestore(&rdev->fence_lock, irq_flags); + return -EDEADLK; } } } diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index c7008b5..e15cb1f 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -154,6 +154,17 @@ void radeon_gem_object_close(struct drm_gem_object *obj, radeon_bo_unreserve(rbo); }

+static int radeon_gem_handle_lockup(struct radeon_device *rdev, int r) +{ + if (r == -EDEADLK) { + radeon_mutex_lock(&rdev->cs_mutex); + r = radeon_gpu_reset(rdev); + if (!r) + r = -EAGAIN; + radeon_mutex_unlock(&rdev->cs_mutex); + } + return r; +}

/* * GEM ioctls. @@ -210,12 +221,14 @@ int radeon_gem_create_ioctl(struct drm_device *dev, void *data, args->initial_domain, false, false, &gobj); if (r) { + r = radeon_gem_handle_lockup(rdev, r); return r; } r = drm_gem_handle_create(filp, gobj, &handle); /* drop reference from allocate - handle holds it now */ drm_gem_object_unreference_unlocked(gobj); if (r) { + r = radeon_gem_handle_lockup(rdev, r); return r; } args->handle = handle; @@ -245,6 +258,7 @@ int radeon_gem_set_domain_ioctl(struct drm_device *dev, void *data, r = radeon_gem_set_domain(gobj, args->read_domains, args->write_domain);

drm_gem_object_unreference_unlocked(gobj); + r = radeon_gem_handle_lockup(robj->rdev, r); return r; }

@@ -301,6 +315,7 @@ int radeon_gem_busy_ioctl(struct drm_device *dev, void *data, break; } drm_gem_object_unreference_unlocked(gobj); + r = radeon_gem_handle_lockup(robj->rdev, r); return r; }

@@ -322,6 +337,7 @@ int radeon_gem_wait_idle_ioctl(struct drm_device *dev, void *data, if (robj->rdev->asic->ioctl_wait_idle) robj->rdev->asic->ioctl_wait_idle(robj->rdev, robj); drm_gem_object_unreference_unlocked(gobj); + r = radeon_gem_handle_lockup(robj->rdev, r); return r; }

From: Christian König deathsimple@vodafone.de

It isn't chipset specific, so it makes no sense to have that inside r100.c.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/evergreen.c | 5 +-- drivers/gpu/drm/radeon/ni.c | 5 +-- drivers/gpu/drm/radeon/r100.c | 57 +-------------------------------- drivers/gpu/drm/radeon/r300.c | 4 +- drivers/gpu/drm/radeon/r600.c | 10 +----- drivers/gpu/drm/radeon/radeon.h | 16 ++------- drivers/gpu/drm/radeon/radeon_asic.h | 5 --- drivers/gpu/drm/radeon/radeon_ring.c | 53 +++++++++++++++++++++++++++++++ drivers/gpu/drm/radeon/si.c | 5 +-- 9 files changed, 69 insertions(+), 91 deletions(-)

diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index a76389c..353d10a 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -2424,7 +2424,6 @@ bool evergreen_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin u32 srbm_status; u32 grbm_status; u32 grbm_status_se0, grbm_status_se1; - struct r100_gpu_lockup *lockup = &rdev->config.evergreen.lockup; int r;

srbm_status = RREG32(SRBM_STATUS); @@ -2432,7 +2431,7 @@ bool evergreen_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin grbm_status_se0 = RREG32(GRBM_STATUS_SE0); grbm_status_se1 = RREG32(GRBM_STATUS_SE1); if (!(grbm_status & GUI_ACTIVE)) { - r100_gpu_lockup_update(lockup, ring); + radeon_ring_lockup_update(ring); return false; } /* force CP activities */ @@ -2444,7 +2443,7 @@ bool evergreen_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin radeon_ring_unlock_commit(rdev, ring); } ring->rptr = RREG32(CP_RB_RPTR); - return r100_gpu_cp_is_lockup(rdev, lockup, ring); + return radeon_ring_test_lockup(rdev, ring); }

static int evergreen_gpu_soft_reset(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index c0b0956..4327b32 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1397,7 +1397,6 @@ bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) u32 srbm_status; u32 grbm_status; u32 grbm_status_se0, grbm_status_se1; - struct r100_gpu_lockup *lockup = &rdev->config.cayman.lockup; int r;

srbm_status = RREG32(SRBM_STATUS); @@ -1405,7 +1404,7 @@ bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) grbm_status_se0 = RREG32(GRBM_STATUS_SE0); grbm_status_se1 = RREG32(GRBM_STATUS_SE1); if (!(grbm_status & GUI_ACTIVE)) { - r100_gpu_lockup_update(lockup, ring); + radeon_ring_lockup_update(ring); return false; } /* force CP activities */ @@ -1418,7 +1417,7 @@ bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) } /* XXX deal with CP0,1,2 */ ring->rptr = RREG32(ring->rptr_reg); - return r100_gpu_cp_is_lockup(rdev, lockup, ring); + return radeon_ring_test_lockup(rdev, ring); }

static int cayman_gpu_soft_reset(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index ac53bd8..cd00909 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -2155,59 +2155,6 @@ int r100_mc_wait_for_idle(struct radeon_device *rdev) return -1; }

-void r100_gpu_lockup_update(struct r100_gpu_lockup *lockup, struct radeon_ring *ring) -{ - lockup->last_cp_rptr = ring->rptr; - lockup->last_jiffies = jiffies; -} - -/** - * r100_gpu_cp_is_lockup() - check if CP is lockup by recording information - * @rdev: radeon device structure - * @lockup: r100_gpu_lockup structure holding CP lockup tracking informations - * @cp: radeon_cp structure holding CP information - * - * We don't need to initialize the lockup tracking information as we will either - * have CP rptr to a different value of jiffies wrap around which will force - * initialization of the lockup tracking informations. - * - * A possible false positivie is if we get call after while and last_cp_rptr == - * the current CP rptr, even if it's unlikely it might happen. To avoid this - * if the elapsed time since last call is bigger than 2 second than we return - * false and update the tracking information. Due to this the caller must call - * r100_gpu_cp_is_lockup several time in less than 2sec for lockup to be reported - * the fencing code should be cautious about that. - * - * Caller should write to the ring to force CP to do something so we don't get - * false positive when CP is just gived nothing to do. - * - **/ -bool r100_gpu_cp_is_lockup(struct radeon_device *rdev, struct r100_gpu_lockup *lockup, struct radeon_ring *ring) -{ - unsigned long cjiffies, elapsed; - - cjiffies = jiffies; - if (!time_after(cjiffies, lockup->last_jiffies)) { - /* likely a wrap around */ - lockup->last_cp_rptr = ring->rptr; - lockup->last_jiffies = jiffies; - return false; - } - if (ring->rptr != lockup->last_cp_rptr) { - /* CP is still working no lockup */ - lockup->last_cp_rptr = ring->rptr; - lockup->last_jiffies = jiffies; - return false; - } - elapsed = jiffies_to_msecs(cjiffies - lockup->last_jiffies); - if (elapsed >= 10000) { - dev_err(rdev->dev, "GPU lockup CP stall for more than %lumsec\n", elapsed); - return true; - } - /* give a chance to the GPU ... */ - return false; -} - bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) { u32 rbbm_status; @@ -2215,7 +2162,7 @@ bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)

rbbm_status = RREG32(R_000E40_RBBM_STATUS); if (!G_000E40_GUI_ACTIVE(rbbm_status)) { - r100_gpu_lockup_update(&rdev->config.r100.lockup, ring); + radeon_ring_lockup_update(ring); return false; } /* force CP activities */ @@ -2227,7 +2174,7 @@ bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_unlock_commit(rdev, ring); } ring->rptr = RREG32(ring->rptr_reg); - return r100_gpu_cp_is_lockup(rdev, &rdev->config.r100.lockup, ring); + return radeon_ring_test_lockup(rdev, ring); }

void r100_bm_disable(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index 26e0db8..e207664 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -384,7 +384,7 @@ bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)

rbbm_status = RREG32(R_000E40_RBBM_STATUS); if (!G_000E40_GUI_ACTIVE(rbbm_status)) { - r100_gpu_lockup_update(&rdev->config.r300.lockup, ring); + radeon_ring_lockup_update(ring); return false; } /* force CP activities */ @@ -396,7 +396,7 @@ bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_unlock_commit(rdev, ring); } ring->rptr = RREG32(RADEON_CP_RB_RPTR); - return r100_gpu_cp_is_lockup(rdev, &rdev->config.r300.lockup, ring); + return radeon_ring_test_lockup(rdev, ring); }

int r300_asic_reset(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index a252160..48f6fae 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -1350,19 +1350,13 @@ bool r600_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) u32 srbm_status; u32 grbm_status; u32 grbm_status2; - struct r100_gpu_lockup *lockup; int r;

- if (rdev->family >= CHIP_RV770) - lockup = &rdev->config.rv770.lockup; - else - lockup = &rdev->config.r600.lockup; - srbm_status = RREG32(R_000E50_SRBM_STATUS); grbm_status = RREG32(R_008010_GRBM_STATUS); grbm_status2 = RREG32(R_008014_GRBM_STATUS2); if (!G_008010_GUI_ACTIVE(grbm_status)) { - r100_gpu_lockup_update(lockup, ring); + radeon_ring_lockup_update(ring); return false; } /* force CP activities */ @@ -1374,7 +1368,7 @@ bool r600_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_unlock_commit(rdev, ring); } ring->rptr = RREG32(ring->rptr_reg); - return r100_gpu_cp_is_lockup(rdev, lockup, ring); + return radeon_ring_test_lockup(rdev, ring); }

int r600_asic_reset(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index d573356..a5d707d 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -596,6 +596,8 @@ struct radeon_ring { unsigned ring_size; unsigned ring_free_dw; int count_dw; + unsigned long last_activity; + unsigned last_rptr; uint64_t gpu_addr; uint32_t align_mask; uint32_t ptr_mask; @@ -739,6 +741,8 @@ void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *cp); void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_ring *cp); void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_ring *cp); int radeon_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); +void radeon_ring_lockup_update(struct radeon_ring *ring); +bool radeon_ring_test_lockup(struct radeon_device *rdev, struct radeon_ring *ring); int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ring_size, unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg, u32 ptr_reg_shift, u32 ptr_reg_mask, u32 nop); @@ -1188,16 +1192,10 @@ struct radeon_asic { /* * Asic structures */ -struct r100_gpu_lockup { - unsigned long last_jiffies; - u32 last_cp_rptr; -}; - struct r100_asic { const unsigned *reg_safe_bm; unsigned reg_safe_bm_size; u32 hdp_cntl; - struct r100_gpu_lockup lockup; };

struct r300_asic { @@ -1205,7 +1203,6 @@ struct r300_asic { unsigned reg_safe_bm_size; u32 resync_scratch; u32 hdp_cntl; - struct r100_gpu_lockup lockup; };

struct r600_asic { @@ -1227,7 +1224,6 @@ struct r600_asic { unsigned tiling_group_size; unsigned tile_config; unsigned backend_map; - struct r100_gpu_lockup lockup; };

struct rv770_asic { @@ -1253,7 +1249,6 @@ struct rv770_asic { unsigned tiling_group_size; unsigned tile_config; unsigned backend_map; - struct r100_gpu_lockup lockup; };

struct evergreen_asic { @@ -1280,7 +1275,6 @@ struct evergreen_asic { unsigned tiling_group_size; unsigned tile_config; unsigned backend_map; - struct r100_gpu_lockup lockup; };

struct cayman_asic { @@ -1319,7 +1313,6 @@ struct cayman_asic { unsigned multi_gpu_tile_size;

unsigned tile_config; - struct r100_gpu_lockup lockup; };

struct si_asic { @@ -1350,7 +1343,6 @@ struct si_asic { unsigned multi_gpu_tile_size;

unsigned tile_config; - struct r100_gpu_lockup lockup; };

union radeon_asic_config { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 3d9f9f1..d3ec546 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -103,11 +103,6 @@ int r100_pci_gart_enable(struct radeon_device *rdev); void r100_pci_gart_disable(struct radeon_device *rdev); int r100_debugfs_mc_info_init(struct radeon_device *rdev); int r100_gui_wait_for_idle(struct radeon_device *rdev); -void r100_gpu_lockup_update(struct r100_gpu_lockup *lockup, - struct radeon_ring *cp); -bool r100_gpu_cp_is_lockup(struct radeon_device *rdev, - struct r100_gpu_lockup *lockup, - struct radeon_ring *cp); void r100_ib_fini(struct radeon_device *rdev); int r100_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); void r100_irq_disable(struct radeon_device *rdev); diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 7031d69..b34575f 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -352,6 +352,59 @@ void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_ring *rin mutex_unlock(&ring->mutex); }

+void radeon_ring_lockup_update(struct radeon_ring *ring) +{ + ring->last_rptr = ring->rptr; + ring->last_activity = jiffies; +} + +/** + * radeon_ring_test_lockup() - check if ring is lockedup by recording information + * @rdev: radeon device structure + * @ring: radeon_ring structure holding ring information + * + * We don't need to initialize the lockup tracking information as we will either + * have CP rptr to a different value of jiffies wrap around which will force + * initialization of the lockup tracking informations. + * + * A possible false positivie is if we get call after while and last_cp_rptr == + * the current CP rptr, even if it's unlikely it might happen. To avoid this + * if the elapsed time since last call is bigger than 2 second than we return + * false and update the tracking information. Due to this the caller must call + * radeon_ring_test_lockup several time in less than 2sec for lockup to be reported + * the fencing code should be cautious about that. + * + * Caller should write to the ring to force CP to do something so we don't get + * false positive when CP is just gived nothing to do. + * + **/ +bool radeon_ring_test_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + unsigned long cjiffies, elapsed; + uint32_t rptr; + + cjiffies = jiffies; + if (!time_after(cjiffies, ring->last_activity)) { + /* likely a wrap around */ + radeon_ring_lockup_update(ring); + return false; + } + rptr = RREG32(ring->rptr_reg); + ring->rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + if (ring->rptr != ring->last_rptr) { + /* CP is still working no lockup */ + radeon_ring_lockup_update(ring); + return false; + } + elapsed = jiffies_to_msecs(cjiffies - ring->last_activity); + if (elapsed >= 10000) { + dev_err(rdev->dev, "GPU lockup CP stall for more than %lumsec\n", elapsed); + return true; + } + /* give a chance to the GPU ... */ + return false; +} + int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsigned ring_size, unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg, u32 ptr_reg_shift, u32 ptr_reg_mask, u32 nop) diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index f05d37a..e76d8b1 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -2217,7 +2217,6 @@ bool si_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) u32 srbm_status; u32 grbm_status, grbm_status2; u32 grbm_status_se0, grbm_status_se1; - struct r100_gpu_lockup *lockup = &rdev->config.si.lockup; int r;

srbm_status = RREG32(SRBM_STATUS); @@ -2226,7 +2225,7 @@ bool si_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) grbm_status_se0 = RREG32(GRBM_STATUS_SE0); grbm_status_se1 = RREG32(GRBM_STATUS_SE1); if (!(grbm_status & GUI_ACTIVE)) { - r100_gpu_lockup_update(lockup, ring); + radeon_ring_lockup_update(ring); return false; } /* force CP activities */ @@ -2239,7 +2238,7 @@ bool si_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) } /* XXX deal with CP0,1,2 */ ring->rptr = RREG32(ring->rptr_reg); - return r100_gpu_cp_is_lockup(rdev, lockup, ring); + return radeon_ring_test_lockup(rdev, ring); }

static int si_gpu_soft_reset(struct radeon_device *rdev)

From: Christian König deathsimple@vodafone.de

Fixing just another deadlock problem with gpu reset tests.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/radeon_ring.c | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 65286b7..4069b01 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -302,7 +302,9 @@ int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_ring *ring, unsi if (ndw < ring->ring_free_dw) { break; } + mutex_unlock(&ring->mutex); r = radeon_fence_wait_next(rdev, radeon_ring_index(rdev, ring)); + mutex_lock(&ring->mutex); if (r) return r; }

From: Christian König deathsimple@vodafone.de

Since it is now identical to r100_gpu_is_lockup.

Signed-off-by: Christian König deathsimple@vodafone.de Reviewed-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/r300.c | 14 -------------- drivers/gpu/drm/radeon/radeon_asic.c | 16 ++++++++-------- drivers/gpu/drm/radeon/radeon_asic.h | 1 - 3 files changed, 8 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index 04ec269..6419a59 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -377,20 +377,6 @@ void r300_gpu_init(struct radeon_device *rdev) rdev->num_gb_pipes, rdev->num_z_pipes); }

-bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) -{ - u32 rbbm_status; - - rbbm_status = RREG32(R_000E40_RBBM_STATUS); - if (!G_000E40_GUI_ACTIVE(rbbm_status)) { - radeon_ring_lockup_update(ring); - return false; - } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); - return radeon_ring_test_lockup(rdev, ring); -} - int r300_asic_reset(struct radeon_device *rdev) { struct r100_mc_save save; diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index 958b9ea..5e5694e 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -299,7 +299,7 @@ static struct radeon_asic r300_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -373,7 +373,7 @@ static struct radeon_asic r300_asic_pcie = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -447,7 +447,7 @@ static struct radeon_asic r420_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -521,7 +521,7 @@ static struct radeon_asic rs400_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -595,7 +595,7 @@ static struct radeon_asic rs600_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -669,7 +669,7 @@ static struct radeon_asic rs690_asic = { .ring_start = &r300_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -743,7 +743,7 @@ static struct radeon_asic rv515_asic = { .ring_start = &rv515_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { @@ -817,7 +817,7 @@ static struct radeon_asic r520_asic = { .ring_start = &rv515_ring_start, .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, - .is_lockup = &r300_gpu_is_lockup, + .is_lockup = &r100_gpu_is_lockup, } }, .irq = { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index d3ec546..3f7ecbb 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -154,7 +154,6 @@ extern int r300_init(struct radeon_device *rdev); extern void r300_fini(struct radeon_device *rdev); extern int r300_suspend(struct radeon_device *rdev); extern int r300_resume(struct radeon_device *rdev); -extern bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp); extern int r300_asic_reset(struct radeon_device *rdev); extern void r300_ring_start(struct radeon_device *rdev, struct radeon_ring *ring); extern void r300_fence_ring_emit(struct radeon_device *rdev,