Re: [RFC PATCH] fence: dma-buf cross-device synchronization (v12)
Op 12-08-13 17:43, Rob Clark schreef: On Mon, Jul 29, 2013 at 10:05 AM, Maarten Lankhorst maarten.lankho...@canonical.com wrote: A fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A driver must allocate a fence context for each execution ring that can run in parallel. The function for this takes an argument with how many contexts to allocate: + fence_context_alloc() A fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence: + fence_signal() To have a rough approximation whether a fence is fired, call: + fence_is_signaled() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. The one pending on the fence can add an async callback: + fence_add_callback() The callback can optionally be cancelled with: + fence_remove_callback() To wait synchronously, optionally with a timeout: + fence_wait() + fence_wait_timeout() A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = custom_get_fence(...); if ((seqno_fence = to_seqno_fence(fence)) != NULL) { dma_buf *fence_buf = fence-sync_buf; get_dma_buf(fence_buf); ... tell the hw the memory location to wait ... custom_wait_on(fence_buf, fence-seqno_ofs, fence-seqno); } else { /* fall-back to sw sync * / fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. enable_signaling callback is used to provide sw signaling in case a cpu waiter is requested or no compatible hardware signaling could be used. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw-hw signaling path (it can be handled same as sw-sw case), and therefore the fence-ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw-hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. v6: [ Maarten Lankhorst ] I removed dma_fence_cancel_callback and some comments about checking if fence fired or not. This is broken by design. waitqueue_active during destruction is now fatal, since the signaller should be holding a reference in enable_signalling until it signalled the fence. Pass the original dma_fence_cb along, and call __remove_wait in the dma_fence_callback handler, so that no cleanup needs to be performed. v7: [ Maarten Lankhorst ] Set cb-func and only enable sw signaling if fence wasn't signaled yet, for example for hardware fences that may choose to signal blindly. v8: [ Maarten Lankhorst ] Tons of tiny fixes, moved __dma_fence_init to header and fixed include mess. dma-fence.h now includes dma-buf.h All members are now initialized, so kmalloc can be used for allocating a dma-fence. More documentation added. v9: Change compiler bitfields to flags, change return type of enable_signaling to bool. Rework dma_fence_wait. Added dma_fence_is_signaled and dma_fence_wait_timeout. s/dma// and change exports to non GPL. Added fence_is_signaled and
Re: [RFC PATCH] fence: dma-buf cross-device synchronization (v12)
On Thu, Aug 15, 2013 at 7:16 AM, Maarten Lankhorst maarten.lankho...@canonical.com wrote: Op 12-08-13 17:43, Rob Clark schreef: On Mon, Jul 29, 2013 at 10:05 AM, Maarten Lankhorst maarten.lankho...@canonical.com wrote: + [snip] +/** + * fence_add_callback - add a callback to be called when the fence + * is signaled + * @fence: [in]the fence to wait on + * @cb:[in]the callback to register + * @func: [in]the function to call + * @priv: [in]the argument to pass to function + * + * cb will be initialized by fence_add_callback, no initialization + * by the caller is required. Any number of callbacks can be registered + * to a fence, but a callback can only be registered to one fence at a time. + * + * Note that the callback can be called from an atomic context. If + * fence is already signaled, this function will return -ENOENT (and + * *not* call the callback) + * + * Add a software callback to the fence. Same restrictions apply to + * refcount as it does to fence_wait, however the caller doesn't need to + * keep a refcount to fence afterwards: when software access is enabled, + * the creator of the fence is required to keep the fence alive until + * after it signals with fence_signal. The callback itself can be called + * from irq context. + * + */ +int fence_add_callback(struct fence *fence, struct fence_cb *cb, + fence_func_t func, void *priv) +{ + unsigned long flags; + int ret = 0; + bool was_set; + + if (WARN_ON(!fence || !func)) + return -EINVAL; + + if (test_bit(FENCE_FLAG_SIGNALED_BIT, fence-flags)) + return -ENOENT; + + spin_lock_irqsave(fence-lock, flags); + + was_set = test_and_set_bit(FENCE_FLAG_ENABLE_SIGNAL_BIT, fence-flags); + + if (test_bit(FENCE_FLAG_SIGNALED_BIT, fence-flags)) + ret = -ENOENT; + else if (!was_set !fence-ops-enable_signaling(fence)) { + __fence_signal(fence); + ret = -ENOENT; + } + + if (!ret) { + cb-func = func; + cb-priv = priv; + list_add_tail(cb-node, fence-cb_list); since the user is providing the 'struct fence_cb', why not drop the priv func args, and have some cb-initialize macro, ie. INIT_FENCE_CB(foo-fence, cbfxn); and I guess we can just drop priv and let the user embed fence in whatever structure they like. Ie. make it look a bit how work_struct works. I don't mind killing priv. But a INIT_FENCE_CB macro is silly, when all it would do is set cb-func. So passing it as an argument to fence_add_callback is fine, unless you have a better reason to do so. INIT_WORK seems to have a bit more initialization than us, it seems work can be more complicated than callbacks, because the callbacks can only be called once and work can be rescheduled multiple times. yeah, INIT_WORK does more.. although maybe some day we want INIT_FENCE_CB to do more (ie. if we add some debug features to help catch misuse of fence/fence-cb's). And if nothing else, having it look a bit like other constructs that we have in the kernel seems useful. And with my point below, you'd want INIT_FENCE_CB to do a INIT_LIST_HEAD(), so it is (very) slightly more than just setting the fxn ptr. maybe also, if (!list_empty(cb-node) return -EBUSY? I think checking for list_empty(cb-node) is a terrible idea. This is no different from any other list corruption, and it's a programming error. Not a runtime error. :-) I was thinking for crtc and page-flip, embed the fence_cb in the crtc. You should only use the cb once at a time, but in this case you might want to re-use it for the next page flip. Having something to catch cb mis-use in this sort of scenario seems useful. maybe how I am thinking to use fence_cb is not quite what you had in mind. I'm not sure. I was trying to think how I could just directly use fence/fence_cb in msm for everything (imported dmabuf or just regular 'ol gem buffers). cb-node.next/prev may be NULL, which would fail with this check. The contents of cb-node are undefined before fence_add_callback is called. Calling fence_remove_callback on a fence that hasn't been added is undefined too. Calling fence_remove_callback works, but I'm thinking of changing the list_del_init to list_del, which would make calling fence_remove_callback twice a fatal error if CONFIG_DEBUG_LIST is enabled, and a possible memory corruption otherwise. ... + [snip] + +/** + * fence context counter: each execution context should have its own + * fence context, this allows checking if fences belong to the same + * context or not. One device can have multiple separate contexts, + * and they're used if some engine can run independently of another. + */ +extern atomic_t fence_context_counter; context-alloc should not be in the critical path.. I'd think probably
Re: [RFC PATCH] fence: dma-buf cross-device synchronization (v12)
Op 15-08-13 15:14, Rob Clark schreef: On Thu, Aug 15, 2013 at 7:16 AM, Maarten Lankhorst maarten.lankho...@canonical.com wrote: Op 12-08-13 17:43, Rob Clark schreef: On Mon, Jul 29, 2013 at 10:05 AM, Maarten Lankhorst maarten.lankho...@canonical.com wrote: + [snip] +/** + * fence_add_callback - add a callback to be called when the fence + * is signaled + * @fence: [in]the fence to wait on + * @cb:[in]the callback to register + * @func: [in]the function to call + * @priv: [in]the argument to pass to function + * + * cb will be initialized by fence_add_callback, no initialization + * by the caller is required. Any number of callbacks can be registered + * to a fence, but a callback can only be registered to one fence at a time. + * + * Note that the callback can be called from an atomic context. If + * fence is already signaled, this function will return -ENOENT (and + * *not* call the callback) + * + * Add a software callback to the fence. Same restrictions apply to + * refcount as it does to fence_wait, however the caller doesn't need to + * keep a refcount to fence afterwards: when software access is enabled, + * the creator of the fence is required to keep the fence alive until + * after it signals with fence_signal. The callback itself can be called + * from irq context. + * + */ +int fence_add_callback(struct fence *fence, struct fence_cb *cb, + fence_func_t func, void *priv) +{ + unsigned long flags; + int ret = 0; + bool was_set; + + if (WARN_ON(!fence || !func)) + return -EINVAL; + + if (test_bit(FENCE_FLAG_SIGNALED_BIT, fence-flags)) + return -ENOENT; + + spin_lock_irqsave(fence-lock, flags); + + was_set = test_and_set_bit(FENCE_FLAG_ENABLE_SIGNAL_BIT, fence-flags); + + if (test_bit(FENCE_FLAG_SIGNALED_BIT, fence-flags)) + ret = -ENOENT; + else if (!was_set !fence-ops-enable_signaling(fence)) { + __fence_signal(fence); + ret = -ENOENT; + } + + if (!ret) { + cb-func = func; + cb-priv = priv; + list_add_tail(cb-node, fence-cb_list); since the user is providing the 'struct fence_cb', why not drop the priv func args, and have some cb-initialize macro, ie. INIT_FENCE_CB(foo-fence, cbfxn); and I guess we can just drop priv and let the user embed fence in whatever structure they like. Ie. make it look a bit how work_struct works. I don't mind killing priv. But a INIT_FENCE_CB macro is silly, when all it would do is set cb-func. So passing it as an argument to fence_add_callback is fine, unless you have a better reason to do so. INIT_WORK seems to have a bit more initialization than us, it seems work can be more complicated than callbacks, because the callbacks can only be called once and work can be rescheduled multiple times. yeah, INIT_WORK does more.. although maybe some day we want INIT_FENCE_CB to do more (ie. if we add some debug features to help catch misuse of fence/fence-cb's). And if nothing else, having it look a bit like other constructs that we have in the kernel seems useful. And with my point below, you'd want INIT_FENCE_CB to do a INIT_LIST_HEAD(), so it is (very) slightly more than just setting the fxn ptr. I don't think list is a good idea for that. maybe also, if (!list_empty(cb-node) return -EBUSY? I think checking for list_empty(cb-node) is a terrible idea. This is no different from any other list corruption, and it's a programming error. Not a runtime error. :-) I was thinking for crtc and page-flip, embed the fence_cb in the crtc. You should only use the cb once at a time, but in this case you might want to re-use it for the next page flip. Having something to catch cb mis-use in this sort of scenario seems useful. maybe how I am thinking to use fence_cb is not quite what you had in mind. I'm not sure. I was trying to think how I could just directly use fence/fence_cb in msm for everything (imported dmabuf or just regular 'ol gem buffers). cb-node.next/prev may be NULL, which would fail with this check. The contents of cb-node are undefined before fence_add_callback is called. Calling fence_remove_callback on a fence that hasn't been added is undefined too. Calling fence_remove_callback works, but I'm thinking of changing the list_del_init to list_del, which would make calling fence_remove_callback twice a fatal error if CONFIG_DEBUG_LIST is enabled, and a possible memory corruption otherwise. ... + [snip] + +/** + * fence context counter: each execution context should have its own + * fence context, this allows checking if fences belong to the same + * context or not. One device can have multiple separate contexts, + * and they're used if some engine can run independently of another. + */ +extern
Re: [RFC PATCH] fence: dma-buf cross-device synchronization (v12)
On Mon, Jul 29, 2013 at 10:05 AM, Maarten Lankhorst maarten.lankho...@canonical.com wrote: A fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A driver must allocate a fence context for each execution ring that can run in parallel. The function for this takes an argument with how many contexts to allocate: + fence_context_alloc() A fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence: + fence_signal() To have a rough approximation whether a fence is fired, call: + fence_is_signaled() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. The one pending on the fence can add an async callback: + fence_add_callback() The callback can optionally be cancelled with: + fence_remove_callback() To wait synchronously, optionally with a timeout: + fence_wait() + fence_wait_timeout() A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = custom_get_fence(...); if ((seqno_fence = to_seqno_fence(fence)) != NULL) { dma_buf *fence_buf = fence-sync_buf; get_dma_buf(fence_buf); ... tell the hw the memory location to wait ... custom_wait_on(fence_buf, fence-seqno_ofs, fence-seqno); } else { /* fall-back to sw sync * / fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. enable_signaling callback is used to provide sw signaling in case a cpu waiter is requested or no compatible hardware signaling could be used. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw-hw signaling path (it can be handled same as sw-sw case), and therefore the fence-ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw-hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. v6: [ Maarten Lankhorst ] I removed dma_fence_cancel_callback and some comments about checking if fence fired or not. This is broken by design. waitqueue_active during destruction is now fatal, since the signaller should be holding a reference in enable_signalling until it signalled the fence. Pass the original dma_fence_cb along, and call __remove_wait in the dma_fence_callback handler, so that no cleanup needs to be performed. v7: [ Maarten Lankhorst ] Set cb-func and only enable sw signaling if fence wasn't signaled yet, for example for hardware fences that may choose to signal blindly. v8: [ Maarten Lankhorst ] Tons of tiny fixes, moved __dma_fence_init to header and fixed include mess. dma-fence.h now includes dma-buf.h All members are now initialized, so kmalloc can be used for allocating a dma-fence. More documentation added. v9: Change compiler bitfields to flags, change return type of enable_signaling to bool. Rework dma_fence_wait. Added dma_fence_is_signaled and dma_fence_wait_timeout. s/dma// and change exports to non GPL. Added fence_is_signaled and fence_enable_sw_signaling calls, add ability to override
[RFC PATCH] fence: dma-buf cross-device synchronization (v12)
A fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A driver must allocate a fence context for each execution ring that can run in parallel. The function for this takes an argument with how many contexts to allocate: + fence_context_alloc() A fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence: + fence_signal() To have a rough approximation whether a fence is fired, call: + fence_is_signaled() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. The one pending on the fence can add an async callback: + fence_add_callback() The callback can optionally be cancelled with: + fence_remove_callback() To wait synchronously, optionally with a timeout: + fence_wait() + fence_wait_timeout() A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = custom_get_fence(...); if ((seqno_fence = to_seqno_fence(fence)) != NULL) { dma_buf *fence_buf = fence-sync_buf; get_dma_buf(fence_buf); ... tell the hw the memory location to wait ... custom_wait_on(fence_buf, fence-seqno_ofs, fence-seqno); } else { /* fall-back to sw sync * / fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. enable_signaling callback is used to provide sw signaling in case a cpu waiter is requested or no compatible hardware signaling could be used. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw-hw signaling path (it can be handled same as sw-sw case), and therefore the fence-ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw-hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. v6: [ Maarten Lankhorst ] I removed dma_fence_cancel_callback and some comments about checking if fence fired or not. This is broken by design. waitqueue_active during destruction is now fatal, since the signaller should be holding a reference in enable_signalling until it signalled the fence. Pass the original dma_fence_cb along, and call __remove_wait in the dma_fence_callback handler, so that no cleanup needs to be performed. v7: [ Maarten Lankhorst ] Set cb-func and only enable sw signaling if fence wasn't signaled yet, for example for hardware fences that may choose to signal blindly. v8: [ Maarten Lankhorst ] Tons of tiny fixes, moved __dma_fence_init to header and fixed include mess. dma-fence.h now includes dma-buf.h All members are now initialized, so kmalloc can be used for allocating a dma-fence. More documentation added. v9: Change compiler bitfields to flags, change return type of enable_signaling to bool. Rework dma_fence_wait. Added dma_fence_is_signaled and dma_fence_wait_timeout. s/dma// and change exports to non GPL. Added fence_is_signaled and fence_enable_sw_signaling calls, add ability to override default wait operation. v10: remove event_queue, use a custom list, export try_to_wake_up from scheduler. Remove fence lock and use a global spinlock instead, this should