linux_dsm_epyc7002/drivers/gpu/drm/i915/intel_ringbuffer.h

574 lines
19 KiB
C
Raw Normal View History

#ifndef _INTEL_RINGBUFFER_H_
#define _INTEL_RINGBUFFER_H_
#include <linux/hashtable.h>
#include "i915_gem_batch_pool.h"
#define I915_CMD_HASH_ORDER 9
/* Early gen2 devices have a cacheline of just 32 bytes, using 64 is overkill,
* but keeps the logic simple. Indeed, the whole purpose of this macro is just
* to give some inclination as to some of the magic values used in the various
* workarounds!
*/
#define CACHELINE_BYTES 64
drm/i915/gen8: Add infrastructure to initialize WA batch buffers Some of the WA are to be applied during context save but before restore and some at the end of context save/restore but before executing the instructions in the ring, WA batch buffers are created for this purpose and these WA cannot be applied using normal means. Each context has two registers to load the offsets of these batch buffers. If they are non-zero, HW understands that it need to execute these batches. v1: In this version two separate ring_buffer objects were used to load WA instructions for indirect and per context batch buffers and they were part of every context. v2: Chris suggested to include additional page in context and use it to load these WA instead of creating separate objects. This will simplify lot of things as we need not explicity pin/unpin them. Thomas Daniel further pointed that GuC is planning to use a similar setup to share data between GuC and driver and WA batch buffers can probably share that page. However after discussions with Dave who is implementing GuC changes, he suggested to use an independent page for the reasons - GuC area might grow and these WA are initialized only once and are not changed afterwards so we can share them share across all contexts. The page is updated with WA during render ring init. This has an advantage of not adding more special cases to default_context. We don't know upfront the number of WA we will applying using these batch buffers. For this reason the size was fixed earlier but it is not a good idea. To fix this, the functions that load instructions are modified to report the no of commands inserted and the size is now calculated after the batch is updated. A macro is introduced to add commands to these batch buffers which also checks for overflow and returns error. We have a full page dedicated for these WA so that should be sufficient for good number of WA, anything more means we have major issues. The list for Gen8 is small, same for Gen9 also, maybe few more gets added going forward but not close to filling entire page. Chris suggested a two-pass approach but we agreed to go with single page setup as it is a one-off routine and simpler code wins. One additional option is offset field which is helpful if we would like to have multiple batches at different offsets within the page and select them based on some criteria. This is not a requirement at this point but could help in future (Dave). Chris provided some helpful macros and suggestions which further simplified the code, they will also help in reducing code duplication when WA for other Gen are added. Add detailed comments explaining restrictions. Use do {} while(0) for wa_ctx_emit() macro. (Many thanks to Chris, Dave and Thomas for their reviews and inputs) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Dave Gordon <david.s.gordon@intel.com> Signed-off-by: Rafael Barbalho <rafael.barbalho@intel.com> Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-20 01:07:01 +07:00
#define CACHELINE_DWORDS (CACHELINE_BYTES / sizeof(uint32_t))
/*
* Gen2 BSpec "1. Programming Environment" / 1.4.4.6 "Ring Buffer Use"
* Gen3 BSpec "vol1c Memory Interface Functions" / 2.3.4.5 "Ring Buffer Use"
* Gen4+ BSpec "vol1c Memory Interface and Command Stream" / 5.3.4.5 "Ring Buffer Use"
*
* "If the Ring Buffer Head Pointer and the Tail Pointer are on the same
* cacheline, the Head Pointer must not be greater than the Tail
* Pointer."
*/
#define I915_RING_FREE_SPACE 64
struct intel_hw_status_page {
u32 *page_addr;
unsigned int gfx_addr;
struct drm_i915_gem_object *obj;
};
#define I915_READ_TAIL(ring) I915_READ(RING_TAIL((ring)->mmio_base))
#define I915_WRITE_TAIL(ring, val) I915_WRITE(RING_TAIL((ring)->mmio_base), val)
#define I915_READ_START(ring) I915_READ(RING_START((ring)->mmio_base))
#define I915_WRITE_START(ring, val) I915_WRITE(RING_START((ring)->mmio_base), val)
#define I915_READ_HEAD(ring) I915_READ(RING_HEAD((ring)->mmio_base))
#define I915_WRITE_HEAD(ring, val) I915_WRITE(RING_HEAD((ring)->mmio_base), val)
#define I915_READ_CTL(ring) I915_READ(RING_CTL((ring)->mmio_base))
#define I915_WRITE_CTL(ring, val) I915_WRITE(RING_CTL((ring)->mmio_base), val)
#define I915_READ_IMR(ring) I915_READ(RING_IMR((ring)->mmio_base))
#define I915_WRITE_IMR(ring, val) I915_WRITE(RING_IMR((ring)->mmio_base), val)
#define I915_READ_MODE(ring) I915_READ(RING_MI_MODE((ring)->mmio_base))
#define I915_WRITE_MODE(ring, val) I915_WRITE(RING_MI_MODE((ring)->mmio_base), val)
/* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to
* do the writes, and that must have qw aligned offsets, simply pretend it's 8b.
*/
#define gen8_semaphore_seqno_size sizeof(uint64_t)
#define GEN8_SEMAPHORE_OFFSET(__from, __to) \
(((__from) * I915_NUM_ENGINES + (__to)) * gen8_semaphore_seqno_size)
#define GEN8_SIGNAL_OFFSET(__ring, to) \
(i915_gem_obj_ggtt_offset(dev_priv->semaphore_obj) + \
GEN8_SEMAPHORE_OFFSET((__ring)->id, (to)))
#define GEN8_WAIT_OFFSET(__ring, from) \
(i915_gem_obj_ggtt_offset(dev_priv->semaphore_obj) + \
GEN8_SEMAPHORE_OFFSET(from, (__ring)->id))
enum intel_ring_hangcheck_action {
HANGCHECK_IDLE = 0,
HANGCHECK_WAIT,
HANGCHECK_ACTIVE,
HANGCHECK_KICK,
HANGCHECK_HUNG,
};
#define HANGCHECK_SCORE_RING_HUNG 31
struct intel_ring_hangcheck {
u64 acthd;
unsigned long user_interrupts;
u32 seqno;
drm/i915: detect hang using per ring hangcheck_score Keep track of ring seqno progress and if there are no progress detected, declare hang. Use actual head (acthd) to distinguish between ring stuck and batchbuffer looping situation. Stuck ring will be kicked to trigger progress. This commit adds a hard limit for batchbuffer completion time. If batchbuffer completion time is more than 4.5 seconds, the gpu will be declared hung. Review comment from Ben which nicely clarifies the semantic change: "Maybe I'm just stating the functional changes of the patch, but in case they were unintended here is what I see as potential issues: 1. "If ring B is waiting on ring A via semaphore, and ring A is making progress, albeit slowly - the hangcheck will fire. The check will determine that A is moving, however ring B will appear hung because the ACTHD doesn't move. I honestly can't say if that's actually a realistic problem to hit it probably implies the timeout value is too low. 2. "There's also another corner case on the kick. If the seqno = 2 (though not stuck), and on the 3rd hangcheck, the ring is stuck, and we try to kick it... we don't actually try to find out if the kick helped" v2: use atchd to detect stuck ring from loop (Ben Widawsky) v3: Use acthd to check when ring needs kicking. Declare hang on third time in order to give time for kick_ring to take effect. v4: Update commit msg Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> Reviewed-by: Ben Widawsky <ben@bwidawsk.net> [danvet: Paste in Ben's review comment.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-05-30 13:04:29 +07:00
int score;
enum intel_ring_hangcheck_action action;
int deadlock;
u32 instdone[I915_NUM_INSTDONE_REG];
};
struct intel_ringbuffer {
struct drm_i915_gem_object *obj;
void __iomem *virtual_start;
struct i915_vma *vma;
struct intel_engine_cs *engine;
struct list_head link;
u32 head;
u32 tail;
int space;
int size;
int effective_size;
/** We track the position of the requests in the ring buffer, and
* when each is retired we increment last_retired_head as the GPU
* must have finished processing the request and so we know we
* can advance the ringbuffer up to that position.
*
* last_retired_head is set to -1 after the value is consumed so
* we can detect new retirements.
*/
u32 last_retired_head;
};
struct i915_gem_context;
struct drm_i915_reg_table;
drm/i915/gen8: Add infrastructure to initialize WA batch buffers Some of the WA are to be applied during context save but before restore and some at the end of context save/restore but before executing the instructions in the ring, WA batch buffers are created for this purpose and these WA cannot be applied using normal means. Each context has two registers to load the offsets of these batch buffers. If they are non-zero, HW understands that it need to execute these batches. v1: In this version two separate ring_buffer objects were used to load WA instructions for indirect and per context batch buffers and they were part of every context. v2: Chris suggested to include additional page in context and use it to load these WA instead of creating separate objects. This will simplify lot of things as we need not explicity pin/unpin them. Thomas Daniel further pointed that GuC is planning to use a similar setup to share data between GuC and driver and WA batch buffers can probably share that page. However after discussions with Dave who is implementing GuC changes, he suggested to use an independent page for the reasons - GuC area might grow and these WA are initialized only once and are not changed afterwards so we can share them share across all contexts. The page is updated with WA during render ring init. This has an advantage of not adding more special cases to default_context. We don't know upfront the number of WA we will applying using these batch buffers. For this reason the size was fixed earlier but it is not a good idea. To fix this, the functions that load instructions are modified to report the no of commands inserted and the size is now calculated after the batch is updated. A macro is introduced to add commands to these batch buffers which also checks for overflow and returns error. We have a full page dedicated for these WA so that should be sufficient for good number of WA, anything more means we have major issues. The list for Gen8 is small, same for Gen9 also, maybe few more gets added going forward but not close to filling entire page. Chris suggested a two-pass approach but we agreed to go with single page setup as it is a one-off routine and simpler code wins. One additional option is offset field which is helpful if we would like to have multiple batches at different offsets within the page and select them based on some criteria. This is not a requirement at this point but could help in future (Dave). Chris provided some helpful macros and suggestions which further simplified the code, they will also help in reducing code duplication when WA for other Gen are added. Add detailed comments explaining restrictions. Use do {} while(0) for wa_ctx_emit() macro. (Many thanks to Chris, Dave and Thomas for their reviews and inputs) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Dave Gordon <david.s.gordon@intel.com> Signed-off-by: Rafael Barbalho <rafael.barbalho@intel.com> Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-20 01:07:01 +07:00
/*
* we use a single page to load ctx workarounds so all of these
* values are referred in terms of dwords
*
* struct i915_wa_ctx_bb:
* offset: specifies batch starting position, also helpful in case
* if we want to have multiple batches at different offsets based on
* some criteria. It is not a requirement at the moment but provides
* an option for future use.
* size: size of the batch in DWORDS
*/
struct i915_ctx_workarounds {
struct i915_wa_ctx_bb {
u32 offset;
u32 size;
} indirect_ctx, per_ctx;
struct drm_i915_gem_object *obj;
};
struct drm_i915_gem_request;
struct intel_engine_cs {
struct drm_i915_private *i915;
const char *name;
enum intel_engine_id {
RCS = 0,
BCS,
VCS,
VCS2, /* Keep instances of the same type engine together. */
VECS
} id;
#define I915_NUM_ENGINES 5
#define _VCS(n) (VCS + (n))
unsigned int exec_id;
unsigned int hw_id;
unsigned int guc_id; /* XXX same as hw_id? */
u32 mmio_base;
struct intel_ringbuffer *buffer;
struct list_head buffers;
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
/* Rather than have every client wait upon all user interrupts,
* with the herd waking after every interrupt and each doing the
* heavyweight seqno dance, we delegate the task (of being the
* bottom-half of the user interrupt) to the first client. After
* every interrupt, we wake up one client, who does the heavyweight
* coherent seqno read and either goes back to sleep (if incomplete),
* or wakes up all the completed clients in parallel, before then
* transferring the bottom-half status to the next client in the queue.
*
* Compared to walking the entire list of waiters in a single dedicated
* bottom-half, we reduce the latency of the first waiter by avoiding
* a context switch, but incur additional coherent seqno reads when
* following the chain of request breadcrumbs. Since it is most likely
* that we have a single client waiting on each seqno, then reducing
* the overhead of waking that client is much preferred.
*/
struct intel_breadcrumbs {
struct task_struct *irq_seqno_bh; /* bh for user interrupts */
unsigned long irq_wakeups;
bool irq_posted;
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
spinlock_t lock; /* protects the lists of requests */
struct rb_root waiters; /* sorted by retirement, priority */
struct rb_root signals; /* sorted by retirement */
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
struct intel_wait *first_wait; /* oldest waiter by retirement */
struct task_struct *signaler; /* used for fence signalling */
struct drm_i915_gem_request *first_signal;
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
struct timer_list fake_irq; /* used after a missed interrupt */
bool irq_enabled : 1;
bool rpm_wakelock : 1;
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
} breadcrumbs;
/*
* A pool of objects to use as shadow copies of client batch buffers
* when the command parser is enabled. Prevents the client from
* modifying the batch contents after software parsing.
*/
struct i915_gem_batch_pool batch_pool;
struct intel_hw_status_page status_page;
drm/i915/gen8: Add infrastructure to initialize WA batch buffers Some of the WA are to be applied during context save but before restore and some at the end of context save/restore but before executing the instructions in the ring, WA batch buffers are created for this purpose and these WA cannot be applied using normal means. Each context has two registers to load the offsets of these batch buffers. If they are non-zero, HW understands that it need to execute these batches. v1: In this version two separate ring_buffer objects were used to load WA instructions for indirect and per context batch buffers and they were part of every context. v2: Chris suggested to include additional page in context and use it to load these WA instead of creating separate objects. This will simplify lot of things as we need not explicity pin/unpin them. Thomas Daniel further pointed that GuC is planning to use a similar setup to share data between GuC and driver and WA batch buffers can probably share that page. However after discussions with Dave who is implementing GuC changes, he suggested to use an independent page for the reasons - GuC area might grow and these WA are initialized only once and are not changed afterwards so we can share them share across all contexts. The page is updated with WA during render ring init. This has an advantage of not adding more special cases to default_context. We don't know upfront the number of WA we will applying using these batch buffers. For this reason the size was fixed earlier but it is not a good idea. To fix this, the functions that load instructions are modified to report the no of commands inserted and the size is now calculated after the batch is updated. A macro is introduced to add commands to these batch buffers which also checks for overflow and returns error. We have a full page dedicated for these WA so that should be sufficient for good number of WA, anything more means we have major issues. The list for Gen8 is small, same for Gen9 also, maybe few more gets added going forward but not close to filling entire page. Chris suggested a two-pass approach but we agreed to go with single page setup as it is a one-off routine and simpler code wins. One additional option is offset field which is helpful if we would like to have multiple batches at different offsets within the page and select them based on some criteria. This is not a requirement at this point but could help in future (Dave). Chris provided some helpful macros and suggestions which further simplified the code, they will also help in reducing code duplication when WA for other Gen are added. Add detailed comments explaining restrictions. Use do {} while(0) for wa_ctx_emit() macro. (Many thanks to Chris, Dave and Thomas for their reviews and inputs) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Dave Gordon <david.s.gordon@intel.com> Signed-off-by: Rafael Barbalho <rafael.barbalho@intel.com> Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-20 01:07:01 +07:00
struct i915_ctx_workarounds wa_ctx;
u32 irq_keep_mask; /* always keep these interrupts */
u32 irq_enable_mask; /* bitmask to enable ring interrupt */
void (*irq_enable)(struct intel_engine_cs *ring);
void (*irq_disable)(struct intel_engine_cs *ring);
int (*init_hw)(struct intel_engine_cs *ring);
int (*init_context)(struct drm_i915_gem_request *req);
void (*write_tail)(struct intel_engine_cs *ring,
u32 value);
int __must_check (*flush)(struct drm_i915_gem_request *req,
u32 invalidate_domains,
u32 flush_domains);
int (*add_request)(struct drm_i915_gem_request *req);
/* Some chipsets are not quite as coherent as advertised and need
* an expensive kick to force a true read of the up-to-date seqno.
* However, the up-to-date seqno is not always required and the last
* seen value is good enough. Note that the seqno will always be
* monotonic, even if not coherent.
*/
void (*irq_seqno_barrier)(struct intel_engine_cs *ring);
int (*dispatch_execbuffer)(struct drm_i915_gem_request *req,
u64 offset, u32 length,
unsigned dispatch_flags);
#define I915_DISPATCH_SECURE 0x1
#define I915_DISPATCH_PINNED 0x2
#define I915_DISPATCH_RS 0x4
void (*cleanup)(struct intel_engine_cs *ring);
/* GEN8 signal/wait table - never trust comments!
* signal to signal to signal to signal to signal to
* RCS VCS BCS VECS VCS2
* --------------------------------------------------------------------
* RCS | NOP (0x00) | VCS (0x08) | BCS (0x10) | VECS (0x18) | VCS2 (0x20) |
* |-------------------------------------------------------------------
* VCS | RCS (0x28) | NOP (0x30) | BCS (0x38) | VECS (0x40) | VCS2 (0x48) |
* |-------------------------------------------------------------------
* BCS | RCS (0x50) | VCS (0x58) | NOP (0x60) | VECS (0x68) | VCS2 (0x70) |
* |-------------------------------------------------------------------
* VECS | RCS (0x78) | VCS (0x80) | BCS (0x88) | NOP (0x90) | VCS2 (0x98) |
* |-------------------------------------------------------------------
* VCS2 | RCS (0xa0) | VCS (0xa8) | BCS (0xb0) | VECS (0xb8) | NOP (0xc0) |
* |-------------------------------------------------------------------
*
* Generalization:
* f(x, y) := (x->id * NUM_RINGS * seqno_size) + (seqno_size * y->id)
* ie. transpose of g(x, y)
*
* sync from sync from sync from sync from sync from
* RCS VCS BCS VECS VCS2
* --------------------------------------------------------------------
* RCS | NOP (0x00) | VCS (0x28) | BCS (0x50) | VECS (0x78) | VCS2 (0xa0) |
* |-------------------------------------------------------------------
* VCS | RCS (0x08) | NOP (0x30) | BCS (0x58) | VECS (0x80) | VCS2 (0xa8) |
* |-------------------------------------------------------------------
* BCS | RCS (0x10) | VCS (0x38) | NOP (0x60) | VECS (0x88) | VCS2 (0xb0) |
* |-------------------------------------------------------------------
* VECS | RCS (0x18) | VCS (0x40) | BCS (0x68) | NOP (0x90) | VCS2 (0xb8) |
* |-------------------------------------------------------------------
* VCS2 | RCS (0x20) | VCS (0x48) | BCS (0x70) | VECS (0x98) | NOP (0xc0) |
* |-------------------------------------------------------------------
*
* Generalization:
* g(x, y) := (y->id * NUM_RINGS * seqno_size) + (seqno_size * x->id)
* ie. transpose of f(x, y)
*/
struct {
u32 sync_seqno[I915_NUM_ENGINES-1];
union {
struct {
/* our mbox written by others */
u32 wait[I915_NUM_ENGINES];
/* mboxes this ring signals to */
i915_reg_t signal[I915_NUM_ENGINES];
} mbox;
u64 signal_ggtt[I915_NUM_ENGINES];
};
/* AKA wait() */
int (*sync_to)(struct drm_i915_gem_request *to_req,
struct intel_engine_cs *from,
u32 seqno);
int (*signal)(struct drm_i915_gem_request *signaller_req,
/* num_dwords needed by caller */
unsigned int num_dwords);
} semaphore;
/* Execlists */
drm/i915: Move execlists irq handler to a bottom half Doing a lot of work in the interrupt handler introduces huge latencies to the system as a whole. Most dramatic effect can be seen by running an all engine stress test like igt/gem_exec_nop/all where, when the kernel config is lean enough, the whole system can be brought into multi-second periods of complete non-interactivty. That can look for example like this: NMI watchdog: BUG: soft lockup - CPU#0 stuck for 23s! [kworker/u8:3:143] Modules linked in: [redacted for brevity] CPU: 0 PID: 143 Comm: kworker/u8:3 Tainted: G U L 4.5.0-160321+ #183 Hardware name: Intel Corporation Broadwell Client platform/WhiteTip Mountain 1 Workqueue: i915 gen6_pm_rps_work [i915] task: ffff8800aae88000 ti: ffff8800aae90000 task.ti: ffff8800aae90000 RIP: 0010:[<ffffffff8104a3c2>] [<ffffffff8104a3c2>] __do_softirq+0x72/0x1d0 RSP: 0000:ffff88014f403f38 EFLAGS: 00000206 RAX: ffff8800aae94000 RBX: 0000000000000000 RCX: 00000000000006e0 RDX: 0000000000000020 RSI: 0000000004208060 RDI: 0000000000215d80 RBP: ffff88014f403f80 R08: 0000000b1b42c180 R09: 0000000000000022 R10: 0000000000000004 R11: 00000000ffffffff R12: 000000000000a030 R13: 0000000000000082 R14: ffff8800aa4d0080 R15: 0000000000000082 FS: 0000000000000000(0000) GS:ffff88014f400000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007fa53b90c000 CR3: 0000000001a0a000 CR4: 00000000001406f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Stack: 042080601b33869f ffff8800aae94000 00000000fffc2678 ffff88010000000a 0000000000000000 000000000000a030 0000000000005302 ffff8800aa4d0080 0000000000000206 ffff88014f403f90 ffffffff8104a716 ffff88014f403fa8 Call Trace: <IRQ> [<ffffffff8104a716>] irq_exit+0x86/0x90 [<ffffffff81031e7d>] smp_apic_timer_interrupt+0x3d/0x50 [<ffffffff814f3eac>] apic_timer_interrupt+0x7c/0x90 <EOI> [<ffffffffa01c5b40>] ? gen8_write64+0x1a0/0x1a0 [i915] [<ffffffff814f2b39>] ? _raw_spin_unlock_irqrestore+0x9/0x20 [<ffffffffa01c5c44>] gen8_write32+0x104/0x1a0 [i915] [<ffffffff8132c6a2>] ? n_tty_receive_buf_common+0x372/0xae0 [<ffffffffa017cc9e>] gen6_set_rps_thresholds+0x1be/0x330 [i915] [<ffffffffa017eaf0>] gen6_set_rps+0x70/0x200 [i915] [<ffffffffa0185375>] intel_set_rps+0x25/0x30 [i915] [<ffffffffa01768fd>] gen6_pm_rps_work+0x10d/0x2e0 [i915] [<ffffffff81063852>] ? finish_task_switch+0x72/0x1c0 [<ffffffff8105ab29>] process_one_work+0x139/0x350 [<ffffffff8105b186>] worker_thread+0x126/0x490 [<ffffffff8105b060>] ? rescuer_thread+0x320/0x320 [<ffffffff8105fa64>] kthread+0xc4/0xe0 [<ffffffff8105f9a0>] ? kthread_create_on_node+0x170/0x170 [<ffffffff814f351f>] ret_from_fork+0x3f/0x70 [<ffffffff8105f9a0>] ? kthread_create_on_node+0x170/0x170 I could not explain, or find a code path, which would explain a +20 second lockup, but from some instrumentation it was apparent the interrupts off proportion of time was between 10-25% under heavy load which is quite bad. When a interrupt "cliff" is reached, which was >~320k irq/s on my machine, the whole system goes into a terrible state of the above described multi-second lockups. By moving the GT interrupt handling to a tasklet in a most simple way, the problem above disappears completely. Testing the effect on sytem-wide latencies using igt/gem_syslatency shows the following before this patch: gem_syslatency: cycles=1532739, latency mean=416531.829us max=2499237us gem_syslatency: cycles=1839434, latency mean=1458099.157us max=4998944us gem_syslatency: cycles=1432570, latency mean=2688.451us max=1201185us gem_syslatency: cycles=1533543, latency mean=416520.499us max=2498886us This shows that the unrelated process is experiencing huge delays in its wake-up latency. After the patch the results look like this: gem_syslatency: cycles=808907, latency mean=53.133us max=1640us gem_syslatency: cycles=862154, latency mean=62.778us max=2117us gem_syslatency: cycles=856039, latency mean=58.079us max=2123us gem_syslatency: cycles=841683, latency mean=56.914us max=1667us Showing a huge improvement in the unrelated process wake-up latency. It also shows an approximate halving in the number of total empty batches submitted during the test. This may not be worrying since the test puts the driver under a very unrealistic load with ncpu threads doing empty batch submission to all GPU engines each. Another benefit compared to the hard-irq handling is that now work on all engines can be dispatched in parallel since we can have up to number of CPUs active tasklets. (While previously a single hard-irq would serially dispatch on one engine after another.) More interesting scenario with regards to throughput is "gem_latency -n 100" which shows 25% better throughput and CPU usage, and 14% better dispatch latencies. I did not find any gains or regressions with Synmark2 or GLbench under light testing. More benchmarking is certainly required. v2: * execlists_lock should be taken as spin_lock_bh when queuing work from userspace now. (Chris Wilson) * uncore.lock must be taken with spin_lock_irq when submitting requests since that now runs from either softirq or process context. v3: * Expanded commit message with more testing data; * converted missed locking sites to _bh; * added execlist_lock comment. (Chris Wilson) v4: * Mention dispatch parallelism in commit. (Chris Wilson) * Do not hold uncore.lock over MMIO reads since the block is already serialised per-engine via the tasklet itself. (Chris Wilson) * intel_lrc_irq_handler should be static. (Chris Wilson) * Cancel/sync the tasklet on GPU reset. (Chris Wilson) * Document and WARN that tasklet cannot be active/pending on engine cleanup. (Chris Wilson/Imre Deak) Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Imre Deak <imre.deak@intel.com> Testcase: igt/gem_exec_nop/all Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94350 Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1459768316-6670-1-git-send-email-tvrtko.ursulin@linux.intel.com
2016-04-04 18:11:56 +07:00
struct tasklet_struct irq_tasklet;
spinlock_t execlist_lock; /* used inside tasklet, use spin_lock_bh */
struct list_head execlist_queue;
unsigned int fw_domains;
drm/i915: Execlists small cleanups and micro-optimisations Assorted changes in the areas of code cleanup, reduction of invariant conditional in the interrupt handler and lock contention and MMIO access optimisation. * Remove needless initialization. * Improve cache locality by reorganizing code and/or using branch hints to keep unexpected or error conditions out of line. * Favor busy submit path vs. empty queue. * Less branching in hot-paths. v2: * Avoid mmio reads when possible. (Chris Wilson) * Use natural integer size for csb indices. * Remove useless return value from execlists_update_context. * Extract 32-bit ppgtt PDPs update so it is out of line and shared with two callers. * Grab forcewake across all mmio operations to ease the load on uncore lock and use chepear mmio ops. v3: * Removed some more pointless u8 data types. * Removed unused return from execlists_context_queue. * Commit message updates. v4: * Unclumsify the unqueue if statement. (Chris Wilson) * Hide forcewake from the queuing function. (Chris Wilson) Version 3 now makes the irq handling code path ~20% smaller on 48-bit PPGTT hardware, and a little bit less elsewhere. Hot paths are mostly in-line now and hammering on the uncore spinlock is greatly reduced together with mmio traffic to an extent. Benchmarking with "gem_latency -n 100" (keep submitting batches with 100 nop instruction) shows approximately 4% higher throughput, 2% less CPU time and 22% smaller latencies. This was on a big-core while small-cores could benefit even more. Most likely reason for the improvements are the MMIO optimization and uncore lock traffic reduction. One odd result is with "gem_latency -n 0" (dispatching empty batches) which shows 5% more throughput, 8% less CPU time, 25% better producer and consumer latencies, but 15% higher dispatch latency which is yet unexplained. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1456505912-22286-1-git-send-email-tvrtko.ursulin@linux.intel.com
2016-02-26 23:58:32 +07:00
unsigned int next_context_status_buffer;
unsigned int idle_lite_restore_wa;
bool disable_lite_restore_wa;
u32 ctx_desc_template;
int (*emit_request)(struct drm_i915_gem_request *request);
int (*emit_flush)(struct drm_i915_gem_request *request,
u32 invalidate_domains,
u32 flush_domains);
int (*emit_bb_start)(struct drm_i915_gem_request *req,
u64 offset, unsigned dispatch_flags);
/**
* List of objects currently involved in rendering from the
* ringbuffer.
*
* Includes buffers having the contents of their GPU caches
* flushed, not necessarily primitives. last_read_req
* represents when the rendering involved will be completed.
*
* A reference is held on the buffer while on this list.
*/
struct list_head active_list;
/**
* List of breadcrumbs associated with GPU requests currently
* outstanding.
*/
struct list_head request_list;
drm/i915: Snapshot seqno of most recently submitted request. The hang checker needs to inspect whether or not the ring request list is empty as well as if the given engine has reached or passed the most recently submitted request. The problem with this is that the hang checker cannot grab the struct_mutex, which is required in order to safely inspect requests since requests might be deallocated during inspection. In the past we've had kernel panics due to this very unsynchronized access in the hang checker. One solution to this problem is to not inspect the requests directly since we're only interested in the seqno of the most recently submitted request - not the request itself. Instead the seqno of the most recently submitted request is stored separately, which the hang checker then inspects, circumventing the issue of synchronization from the hang checker entirely. This fixes a regression introduced in commit 44cdd6d219bc64f6810b8ed0023a4d4db9e0fe68 Author: John Harrison <John.C.Harrison@Intel.com> Date: Mon Nov 24 18:49:40 2014 +0000 drm/i915: Convert 'ring_idle()' to use requests not seqnos v2 (Chris Wilson): - Pass current engine seqno to ring_idle() from i915_hangcheck_elapsed() rather than compute it over again. - Remove extra whitespace. Issue: VIZ-5998 Signed-off-by: Tomas Elf <tomas.elf@intel.com> Cc: stable@vger.kernel.org Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: Add regressing commit citation provided by Chris.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-07-09 21:30:57 +07:00
/**
* Seqno of request most recently submitted to request_list.
* Used exclusively by hang checker to avoid grabbing lock while
* inspecting request list.
*/
u32 last_submitted_seqno;
2012-06-14 01:45:19 +07:00
bool gpu_caches_dirty;
struct i915_gem_context *last_context;
struct intel_ring_hangcheck hangcheck;
struct {
struct drm_i915_gem_object *obj;
u32 gtt_offset;
} scratch;
bool needs_cmd_parser;
/*
* Table of commands the command parser needs to know about
* for this ring.
*/
DECLARE_HASHTABLE(cmd_hash, I915_CMD_HASH_ORDER);
/*
* Table of registers allowed in commands that read/write registers.
*/
const struct drm_i915_reg_table *reg_tables;
int reg_table_count;
/*
* Returns the bitmask for the length field of the specified command.
* Return 0 for an unrecognized/invalid command.
*
* If the command parser finds an entry for a command in the ring's
* cmd_tables, it gets the command's length based on the table entry.
* If not, it calls this function to determine the per-ring length field
* encoding for the command (i.e. certain opcode ranges use certain bits
* to encode the command length in the header).
*/
u32 (*get_cmd_length_mask)(u32 cmd_header);
};
static inline bool
intel_engine_initialized(const struct intel_engine_cs *engine)
{
return engine->i915 != NULL;
}
static inline unsigned
intel_engine_flag(const struct intel_engine_cs *engine)
{
return 1 << engine->id;
}
static inline u32
intel_ring_sync_index(struct intel_engine_cs *engine,
struct intel_engine_cs *other)
{
int idx;
/*
* rcs -> 0 = vcs, 1 = bcs, 2 = vecs, 3 = vcs2;
* vcs -> 0 = bcs, 1 = vecs, 2 = vcs2, 3 = rcs;
* bcs -> 0 = vecs, 1 = vcs2. 2 = rcs, 3 = vcs;
* vecs -> 0 = vcs2, 1 = rcs, 2 = vcs, 3 = bcs;
* vcs2 -> 0 = rcs, 1 = vcs, 2 = bcs, 3 = vecs;
*/
idx = (other - engine) - 1;
if (idx < 0)
idx += I915_NUM_ENGINES;
return idx;
}
static inline void
intel_flush_status_page(struct intel_engine_cs *engine, int reg)
{
mb();
clflush(&engine->status_page.page_addr[reg]);
mb();
}
static inline u32
intel_read_status_page(struct intel_engine_cs *engine, int reg)
{
/* Ensure that the compiler doesn't optimize away the load. */
return READ_ONCE(engine->status_page.page_addr[reg]);
}
static inline void
intel_write_status_page(struct intel_engine_cs *engine,
int reg, u32 value)
{
engine->status_page.page_addr[reg] = value;
}
/*
* Reads a dword out of the status page, which is written to from the command
* queue by automatic updates, MI_REPORT_HEAD, MI_STORE_DATA_INDEX, or
* MI_STORE_DATA_IMM.
*
* The following dwords have a reserved meaning:
* 0x00: ISR copy, updated when an ISR bit not set in the HWSTAM changes.
* 0x04: ring 0 head pointer
* 0x05: ring 1 head pointer (915-class)
* 0x06: ring 2 head pointer (915-class)
* 0x10-0x1b: Context status DWords (GM45)
* 0x1f: Last written status offset. (GM45)
* 0x20-0x2f: Reserved (Gen6+)
*
* The area from dword 0x30 to 0x3ff is available for driver usage.
*/
#define I915_GEM_HWS_INDEX 0x30
#define I915_GEM_HWS_INDEX_ADDR (I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT)
#define I915_GEM_HWS_SCRATCH_INDEX 0x40
#define I915_GEM_HWS_SCRATCH_ADDR (I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT)
struct intel_ringbuffer *
intel_engine_create_ringbuffer(struct intel_engine_cs *engine, int size);
int intel_pin_and_map_ringbuffer_obj(struct drm_i915_private *dev_priv,
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 17:28:56 +07:00
struct intel_ringbuffer *ringbuf);
void intel_unpin_ringbuffer_obj(struct intel_ringbuffer *ringbuf);
void intel_ringbuffer_free(struct intel_ringbuffer *ring);
void intel_stop_engine(struct intel_engine_cs *engine);
void intel_cleanup_engine(struct intel_engine_cs *engine);
int intel_ring_alloc_request_extras(struct drm_i915_gem_request *request);
int __must_check intel_ring_begin(struct drm_i915_gem_request *req, int n);
int __must_check intel_ring_cacheline_align(struct drm_i915_gem_request *req);
static inline void intel_ring_emit(struct intel_engine_cs *engine,
u32 data)
{
struct intel_ringbuffer *ringbuf = engine->buffer;
iowrite32(data, ringbuf->virtual_start + ringbuf->tail);
ringbuf->tail += 4;
}
static inline void intel_ring_emit_reg(struct intel_engine_cs *engine,
drm/i915: Type safe register read/write Make I915_READ and I915_WRITE more type safe by wrapping the register offset in a struct. This should eliminate most of the fumbles we've had with misplaced parens. This only takes care of normal mmio registers. We could extend the idea to other register types and define each with its own struct. That way you wouldn't be able to accidentally pass the wrong thing to a specific register access function. The gpio_reg setup is probably the ugliest thing left. But I figure I'd just leave it for now, and wait for some divine inspiration to strike before making it nice. As for the generated code, it's actually a bit better sometimes. Eg. looking at i915_irq_handler(), we can see the following change: lea 0x70024(%rdx,%rax,1),%r9d mov $0x1,%edx - movslq %r9d,%r9 - mov %r9,%rsi - mov %r9,-0x58(%rbp) - callq *0xd8(%rbx) + mov %r9d,%esi + mov %r9d,-0x48(%rbp) callq *0xd8(%rbx) So previously gcc thought the register offset might be signed and decided to sign extend it, just in case. The rest appears to be mostly just minor shuffling of instructions. v2: i915_mmio_reg_{offset,equal,valid}() helpers added s/_REG/_MMIO/ in the register defines mo more switch statements left to worry about ring_emit stuff got sorted in a prep patch cmd parser, lrc context and w/a batch buildup also in prep patch vgpu stuff cleaned up and moved to a prep patch all other unrelated changes split out v3: Rebased due to BXT DSI/BLC, MOCS, etc. v4: Rebased due to churn, s/i915_mmio_reg_t/i915_reg_t/ Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1447853606-2751-1-git-send-email-ville.syrjala@linux.intel.com
2015-11-18 20:33:26 +07:00
i915_reg_t reg)
{
intel_ring_emit(engine, i915_mmio_reg_offset(reg));
}
static inline void intel_ring_advance(struct intel_engine_cs *engine)
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-11 04:16:32 +07:00
{
struct intel_ringbuffer *ringbuf = engine->buffer;
ringbuf->tail &= ringbuf->size - 1;
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-11 04:16:32 +07:00
}
int __intel_ring_space(int head, int tail, int size);
void intel_ring_update_space(struct intel_ringbuffer *ringbuf);
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-11 04:16:32 +07:00
int __must_check intel_engine_idle(struct intel_engine_cs *engine);
void intel_ring_init_seqno(struct intel_engine_cs *engine, u32 seqno);
int intel_ring_flush_all_caches(struct drm_i915_gem_request *req);
int intel_ring_invalidate_all_caches(struct drm_i915_gem_request *req);
int intel_init_pipe_control(struct intel_engine_cs *engine, int size);
void intel_fini_pipe_control(struct intel_engine_cs *engine);
int intel_init_render_ring_buffer(struct drm_device *dev);
int intel_init_bsd_ring_buffer(struct drm_device *dev);
int intel_init_bsd2_ring_buffer(struct drm_device *dev);
int intel_init_blt_ring_buffer(struct drm_device *dev);
int intel_init_vebox_ring_buffer(struct drm_device *dev);
u64 intel_ring_get_active_head(struct intel_engine_cs *engine);
static inline u32 intel_engine_get_seqno(struct intel_engine_cs *engine)
{
return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
}
int init_workarounds_ring(struct intel_engine_cs *engine);
static inline u32 intel_ring_get_tail(struct intel_ringbuffer *ringbuf)
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 18:25:36 +07:00
{
return ringbuf->tail;
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 18:25:36 +07:00
}
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 19:10:09 +07:00
/*
* Arbitrary size for largest possible 'add request' sequence. The code paths
* are complex and variable. Empirical measurement shows that the worst case
* is BDW at 192 bytes (6 + 6 + 36 dwords), then ILK at 136 bytes. However,
* we need to allocate double the largest single packet within that emission
* to account for tail wraparound (so 6 + 6 + 72 dwords for BDW).
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 19:10:09 +07:00
*/
#define MIN_SPACE_FOR_ADD_REQUEST 336
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 19:10:09 +07:00
static inline u32 intel_hws_seqno_address(struct intel_engine_cs *engine)
{
return engine->status_page.gfx_addr + I915_GEM_HWS_INDEX_ADDR;
}
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
/* intel_breadcrumbs.c -- user interrupt bottom-half for waiters */
struct intel_wait {
struct rb_node node;
struct task_struct *tsk;
u32 seqno;
};
struct intel_signal_node {
struct rb_node node;
struct intel_wait wait;
};
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
int intel_engine_init_breadcrumbs(struct intel_engine_cs *engine);
static inline void intel_wait_init(struct intel_wait *wait, u32 seqno)
{
wait->tsk = current;
wait->seqno = seqno;
}
static inline bool intel_wait_complete(const struct intel_wait *wait)
{
return RB_EMPTY_NODE(&wait->node);
}
bool intel_engine_add_wait(struct intel_engine_cs *engine,
struct intel_wait *wait);
void intel_engine_remove_wait(struct intel_engine_cs *engine,
struct intel_wait *wait);
void intel_engine_enable_signaling(struct drm_i915_gem_request *request);
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
static inline bool intel_engine_has_waiter(struct intel_engine_cs *engine)
{
return READ_ONCE(engine->breadcrumbs.irq_seqno_bh);
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
}
static inline bool intel_engine_wakeup(struct intel_engine_cs *engine)
{
bool wakeup = false;
struct task_struct *tsk = READ_ONCE(engine->breadcrumbs.irq_seqno_bh);
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
/* Note that for this not to dangerously chase a dangling pointer,
* the caller is responsible for ensure that the task remain valid for
* wake_up_process() i.e. that the RCU grace period cannot expire.
*
* Also note that tsk is likely to be in !TASK_RUNNING state so an
* early test for tsk->state != TASK_RUNNING before wake_up_process()
* is unlikely to be beneficial.
*/
if (tsk)
wakeup = wake_up_process(tsk);
return wakeup;
}
void intel_engine_enable_fake_irq(struct intel_engine_cs *engine);
void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine);
unsigned int intel_kick_waiters(struct drm_i915_private *i915);
unsigned int intel_kick_signalers(struct drm_i915_private *i915);
drm/i915: Slaughter the thundering i915_wait_request herd One particularly stressful scenario consists of many independent tasks all competing for GPU time and waiting upon the results (e.g. realtime transcoding of many, many streams). One bottleneck in particular is that each client waits on its own results, but every client is woken up after every batchbuffer - hence the thunder of hooves as then every client must do its heavyweight dance to read a coherent seqno to see if it is the lucky one. Ideally, we only want one client to wake up after the interrupt and check its request for completion. Since the requests must retire in order, we can select the first client on the oldest request to be woken. Once that client has completed his wait, we can then wake up the next client and so on. However, all clients then incur latency as every process in the chain may be delayed for scheduling - this may also then cause some priority inversion. To reduce the latency, when a client is added or removed from the list, we scan the tree for completed seqno and wake up all the completed waiters in parallel. Using igt/benchmarks/gem_latency, we can demonstrate this effect. The benchmark measures the number of GPU cycles between completion of a batch and the client waking up from a call to wait-ioctl. With many concurrent waiters, with each on a different request, we observe that the wakeup latency before the patch scales nearly linearly with the number of waiters (before external factors kick in making the scaling much worse). After applying the patch, we can see that only the single waiter for the request is being woken up, providing a constant wakeup latency for every operation. However, the situation is not quite as rosy for many waiters on the same request, though to the best of my knowledge this is much less likely in practice. Here, we can observe that the concurrent waiters incur extra latency from being woken up by the solitary bottom-half, rather than directly by the interrupt. This appears to be scheduler induced (having discounted adverse effects from having a rbtree walk/erase in the wakeup path), each additional wake_up_process() costs approximately 1us on big core. Another effect of performing the secondary wakeups from the first bottom-half is the incurred delay this imposes on high priority threads - rather than immediately returning to userspace and leaving the interrupt handler to wake the others. To offset the delay incurred with additional waiters on a request, we could use a hybrid scheme that did a quick read in the interrupt handler and dequeued all the completed waiters (incurring the overhead in the interrupt handler, not the best plan either as we then incur GPU submission latency) but we would still have to wake up the bottom-half every time to do the heavyweight slow read. Or we could only kick the waiters on the seqno with the same priority as the current task (i.e. in the realtime waiter scenario, only it is woken up immediately by the interrupt and simply queues the next waiter before returning to userspace, minimising its delay at the expense of the chain, and also reducing contention on its scheduler runqueue). This is effective at avoid long pauses in the interrupt handler and at avoiding the extra latency in realtime/high-priority waiters. v2: Convert from a kworker per engine into a dedicated kthread for the bottom-half. v3: Rename request members and tweak comments. v4: Use a per-engine spinlock in the breadcrumbs bottom-half. v5: Fix race in locklessly checking waiter status and kicking the task on adding a new waiter. v6: Fix deciding when to force the timer to hide missing interrupts. v7: Move the bottom-half from the kthread to the first client process. v8: Reword a few comments v9: Break the busy loop when the interrupt is unmasked or has fired. v10: Comments, unnecessary churn, better debugging from Tvrtko v11: Wake all completed waiters on removing the current bottom-half to reduce the latency of waking up a herd of clients all waiting on the same request. v12: Rearrange missed-interrupt fault injection so that it works with igt/drv_missed_irq_hang v13: Rename intel_breadcrumb and friends to intel_wait in preparation for signal handling. v14: RCU commentary, assert_spin_locked v15: Hide BUG_ON behind the compiler; report on gem_latency findings. v16: Sort seqno-groups by priority so that first-waiter has the highest task priority (and so avoid priority inversion). v17: Add waiters to post-mortem GPU hang state. v18: Return early for a completed wait after acquiring the spinlock. Avoids adding ourselves to the tree if the is already complete, and skips the awkward question of why we don't do completion wakeups for waits earlier than or equal to ourselves. v19: Prepare for init_breadcrumbs to fail. Later patches may want to allocate during init, so be prepared to propagate back the error code. Testcase: igt/gem_concurrent_blit Testcase: igt/benchmarks/gem_latency Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com> Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Dave Gordon <david.s.gordon@intel.com> Cc: "Goel, Akash" <akash.goel@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18 Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
2016-07-01 23:23:15 +07:00
#endif /* _INTEL_RINGBUFFER_H_ */