DMA Buffer Synchronization Framework ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Inki Dae This document is a guide for device-driver writers describing the DMA buffer synchronization API. This document also describes how to use the API to use buffer synchronization mechanism between DMA and DMA, CPU and DMA, and CPU and CPU. The DMA Buffer synchronization API provides buffer synchronization mechanism; i.e., buffer access control to CPU and DMA, and easy-to-use interfaces for device drivers and user application. And this API can be used for all dma devices using system memory as dma buffer, especially for most ARM based SoCs. Motivation ---------- Buffer synchronization issue between DMA and DMA: Sharing a buffer, a device cannot be aware of when the other device will access the shared buffer: a device may access a buffer containing wrong data if the device accesses the shared buffer while another device is still accessing the shared buffer. Therefore, a user process should have waited for the completion of DMA access by another device before a device tries to access the shared buffer. Buffer synchronization issue between CPU and DMA: A user process should consider that when having to send a buffer, filled by CPU, to a device driver for the device driver to access the buffer as a input buffer while CPU and DMA are sharing the buffer. This means that the user process needs to understand how the device driver is worked. Hence, the conventional mechanism not only makes user application complicated but also incurs performance overhead. Buffer synchronization issue between CPU and CPU: In case that two processes share one buffer; shared with DMA also, they may need some mechanism to allow process B to access the shared buffer after the completion of CPU access by process A. Therefore, process B should have waited for the completion of CPU access by process A using the mechanism before trying to access the shared buffer. What is the best way to solve these buffer synchronization issues? We may need a common object that a device driver and a user process notify the common object of when they try to access a shared buffer. That way we could decide when we have to allow or not to allow for CPU or DMA to access the shared buffer through the common object. If so, what could become the common object? Right, that's a dma-buf[1]. Now we have already been using the dma-buf to share one buffer with other drivers. Basic concept ------------- The mechanism of this framework has the following steps, 1. Register dmabufs to a sync object - A task gets a new sync object and can add one or more dmabufs that the task wants to access. This registering should be performed when a device context or an event context such as a page flip event is created or before CPU accesses a shared buffer. dma_buf_sync_get(a sync object, a dmabuf); 2. Lock a sync object - A task tries to lock all dmabufs added in its own sync object. Basically, the lock mechanism uses ww-mutexes[2] to avoid dead lock issue and for race condition between CPU and CPU, CPU and DMA, and DMA and DMA. Taking a lock means that others cannot access all locked dmabufs until the task that locked the corresponding dmabufs, unlocks all the locked dmabufs. This locking should be performed before DMA or CPU accesses these dmabufs. dma_buf_sync_lock(a sync object); 3. Unlock a sync object - The task unlocks all dmabufs added in its own sync object. The unlock means that the DMA or CPU accesses to the dmabufs have been completed so that others may access them. This unlocking should be performed after DMA or CPU has completed accesses to the dmabufs. dma_buf_sync_unlock(a sync object); 4. Unregister one or all dmabufs from a sync object - A task unregisters the given dmabufs from the sync object. This means that the task dosen't want to lock the dmabufs. The unregistering should be performed after DMA or CPU has completed accesses to the dmabufs or when dma_buf_sync_lock() is failed. dma_buf_sync_put(a sync object, a dmabuf); dma_buf_sync_put_all(a sync object); The described steps may be summarized as: get -> lock -> CPU or DMA access to a buffer/s -> unlock -> put This framework includes the following two features. 1. read (shared) and write (exclusive) locks - A task is required to declare the access type when the task tries to register a dmabuf; READ, WRITE, READ DMA, or WRITE DMA. The below is example codes, struct dmabuf_sync *sync; sync = dmabuf_sync_init(NULL, "test sync"); dmabuf_sync_get(sync, dmabuf, DMA_BUF_ACCESS_R); ... 2. Mandatory resource releasing - a task cannot hold a lock indefinitely. A task may never try to unlock a buffer after taking a lock to the buffer. In this case, a timer handler to the corresponding sync object is called in five (default) seconds and then the timed-out buffer is unlocked by work queue handler to avoid lockups and to enforce resources of the buffer. Access types ------------ DMA_BUF_ACCESS_R - CPU will access a buffer for read. DMA_BUF_ACCESS_W - CPU will access a buffer for read or write. DMA_BUF_ACCESS_DMA_R - DMA will access a buffer for read DMA_BUF_ACCESS_DMA_W - DMA will access a buffer for read or write. Generic user interfaces ----------------------- And this framework includes fcntl[3] and select system calls as interfaces exported to user. As you know, user sees a buffer object as a dma-buf file descriptor. fcntl() call with the file descriptor means to lock some buffer region being managed by the dma-buf object. And select call with the file descriptor means to poll the completion event of CPU or DMA access to the dma-buf. API set ------- bool is_dmabuf_sync_supported(void) - Check if dmabuf sync is supported or not. struct dmabuf_sync *dmabuf_sync_init(const char *name, struct dmabuf_sync_priv_ops *ops, void priv*) - Allocate and initialize a new sync object. The caller can get a new sync object for buffer synchronization. ops is used for device driver to clean up its own sync object. For this, each device driver should implement a free callback. priv is used for device driver to get its device context when free callback is called. void dmabuf_sync_fini(struct dmabuf_sync *sync) - Release all resources to the sync object. int dmabuf_sync_get(struct dmabuf_sync *sync, void *sync_buf, unsigned int type) - Get dmabuf sync object. Internally, this function allocates a dmabuf_sync object and adds a given dmabuf to it, and also takes a reference to the dmabuf. The caller can tie up multiple dmabufs into one sync object by calling this function several times. void dmabuf_sync_put(struct dmabuf_sync *sync, struct dma_buf *dmabuf) - Put dmabuf sync object to a given dmabuf. Internally, this function removes a given dmabuf from a sync object and remove the sync object. At this time, the dmabuf is putted. void dmabuf_sync_put_all(struct dmabuf_sync *sync) - Put dmabuf sync object to dmabufs. Internally, this function removes all dmabufs from a sync object and remove the sync object. At this time, all dmabufs are putted. int dmabuf_sync_lock(struct dmabuf_sync *sync) - Lock all dmabufs added in a sync object. The caller should call this function prior to CPU or DMA access to the dmabufs so that others can not access the dmabufs. Internally, this function avoids dead lock issue with ww-mutexes. int dmabuf_sync_single_lock(struct dma_buf *dmabuf) - Lock a dmabuf. The caller should call this function prior to CPU or DMA access to the dmabuf so that others can not access the dmabuf. int dmabuf_sync_unlock(struct dmabuf_sync *sync) - Unlock all dmabufs added in a sync object. The caller should call this function after CPU or DMA access to the dmabufs is completed so that others can access the dmabufs. void dmabuf_sync_single_unlock(struct dma_buf *dmabuf) - Unlock a dmabuf. The caller should call this function after CPU or DMA access to the dmabuf is completed so that others can access the dmabuf. Tutorial for device driver -------------------------- 1. Allocate and Initialize a sync object: static void xxx_dmabuf_sync_free(void *priv) { struct xxx_context *ctx = priv; if (!ctx) return; ctx->sync = NULL; } ... static struct dmabuf_sync_priv_ops driver_specific_ops = { .free = xxx_dmabuf_sync_free, }; ... struct dmabuf_sync *sync; sync = dmabuf_sync_init("test sync", &driver_specific_ops, ctx); ... 2. Add a dmabuf to the sync object when setting up dma buffer relevant registers: dmabuf_sync_get(sync, dmabuf, DMA_BUF_ACCESS_READ); ... 3. Lock all dmabufs of the sync object before DMA or CPU accesses the dmabufs: dmabuf_sync_lock(sync); ... 4. Now CPU or DMA can access all dmabufs locked in step 3. 5. Unlock all dmabufs added in a sync object after DMA or CPU access to these dmabufs is completed: dmabuf_sync_unlock(sync); And call the following functions to release all resources, dmabuf_sync_put_all(sync); dmabuf_sync_fini(sync); Tutorial for user application ----------------------------- fcntl system call: struct flock filelock; 1. Lock a dma buf: filelock.l_type = F_WRLCK or F_RDLCK; /* lock entire region to the dma buf. */ filelock.lwhence = SEEK_CUR; filelock.l_start = 0; filelock.l_len = 0; fcntl(dmabuf fd, F_SETLKW or F_SETLK, &filelock); ... CPU access to the dma buf 2. Unlock a dma buf: filelock.l_type = F_UNLCK; fcntl(dmabuf fd, F_SETLKW or F_SETLK, &filelock); close(dmabuf fd) call would also unlock the dma buf. And for more detail, please refer to [3] select system call: fd_set wdfs or rdfs; FD_ZERO(&wdfs or &rdfs); FD_SET(fd, &wdfs or &rdfs); select(fd + 1, &rdfs, NULL, NULL, NULL); or select(fd + 1, NULL, &wdfs, NULL, NULL); Every time select system call is called, a caller will wait for the completion of DMA or CPU access to a shared buffer if there is someone accessing the shared buffer. If no anyone then select system call will be returned at once. References: [1] http://lwn.net/Articles/470339/ [2] https://patchwork.kernel.org/patch/2625361/ [3] http://linux.die.net/man/2/fcntl