- 13 Nov, 2022 4 commits
-
-
Akhil R authored
Add dma-channel-mask property in Tegra GPCDMA document. The property would help to specify the channels to be used in kernel and reserve few for the firmware. This was previously achieved by limiting the channel number to 31 in the driver. This is wrong and does not align with the hardware. Correct this and set the max interrupts to 32. Signed-off-by:
Akhil R <akhilrajeev@nvidia.com> Acked-by:
Thierry Reding <treding@nvidia.com> Acked-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Link: https://lore.kernel.org/r/20221110171748.40304-2-akhilrajeev@nvidia.comSigned-off-by:
Vinod Koul <vkoul@kernel.org>
-
Thomas Gleixner authored
Nothing in this file needs anything from linux/msi.h Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Vinod Koul <vkoul@kernel.org> Cc: dmaengine@vger.kernel.org Link: https://lore.kernel.org/r/20221113202428.573536003@linutronix.deSigned-off-by:
-
Konrad Dybcio authored
Document the compatible for GPI DMA controller on SM6375 SoC. Signed-off-by:
Konrad Dybcio <konrad.dybcio@linaro.org> Reviewed-by:
-
Fenghua Yu authored
The crc_val in the completion record should be 64 bits and not 32 bits. Fixes: 4ac823e9 ("dmaengine: idxd: fix delta_rec and crc size field for completion record") Reported-by:
Nirav N Shah <nirav.n.shah@intel.com> Signed-off-by:
Fenghua Yu <fenghua.yu@intel.com> Reviewed-by:
Dave Jiang <dave.jiang@intel.com> Link: https://lore.kernel.org/r/20221111012715.2031481-1-fenghua.yu@intel.comSigned-off-by:
-
- 11 Nov, 2022 18 commits
-
-
Tudor Ambarus authored
Convert the driver to use the core virt-dma. The driver will be easier to maintain as it uses the list handling and the tasklet from virt-dma. With the conversion replace the election of a new transfer in the tasklet with the election of the new transfer in the interrupt handler. With this we have a shorter idle window as we remove the scheduling latency of the tasklet. I chose to do this while doing the conversion to virt-dma, because if I made a prerequisite patch with the new transfer election in the irq handler, I would have to duplicate some virt-dma code in the at_hdmac driver that would end up being removed at the virt-dma conversion anyway. So do this in a single step. Signed-off-by:
Tudor Ambarus <tudor.ambarus@microchip.com> Acked-by:
Nicolas Ferre <nicolas.ferre@microchip.com> Link: https://lore.kernel.org/r/20221025090306.297886-1-tudor.ambarus@microchip.com Link: https://lore.kernel.org/r/20221025090306.297886-33-tudor.ambarus@microchip.comSigned-off-by:
-
Tudor Ambarus authored
The pointer to at_dma engine was never used, remove it. Signed-off-by:
-
Tudor Ambarus authored
"chan_common" was misleading and did not suggest that's actually a struct dma_chan underneath. Rename it so that readers can follow the code easier. One may see some checks when running checkpatch. Those have nothing to do with the rename and will be addressed in a further patch. Signed-off-by:
-
Tudor Ambarus authored
"dma_common" name was misleading and did not suggest that's actually a struct dma_device underneath. Rename it so that readers can follow the code easier. One may see some checks and a warning when running checkpatch. Those have nothing to do with the rename and will be addressed in a further patch. Signed-off-by:
-
Tudor Ambarus authored
Use the bitfield access macros in order to clean and to make the driver easier to read. One will see some "line length exceeds 100 columns" checkpatch warnings. I chose to not introduce new lines for regs descriptions in order to preserve the style of the comments throughout the definitions. Style can be fixed in a further patch. Signed-off-by:
-
Tudor Ambarus authored
Do not expose register definitions, structures and helpers via a .h file because there are used only by at_hdmac.c. Since there are no other users, remove the ambiguity and move all the .h contents to the .c file. One may notice some checkpatch warnings and errors with this move. The move was done "as it was", checkpatch complaints can be fixed in a further patch. Signed-off-by:
-
Tudor Ambarus authored
It's a good practice to set the include entries in alphabetic order. It helps humans to read the code easier. Alphabetic order should also prove that each header is self-contained, i.e. can be included without prerequisites. Signed-off-by:
-
Tudor Ambarus authored
Use pm_ptr() macro to fill at_dma_driver.driver.pm. In case CONFIG_PM is not enabled, the macro will return NULL. When NULL, at_dma_dev_pm_ops will end up being unused, so prepend it with the __maybe_unused attribute. Signed-off-by:
-
Tudor Ambarus authored
Clocks that are get with this method will be automatically put on driver detach. Use devm_clk_get() and simplify the error handling. Signed-off-by:
-
Tudor Ambarus authored
Use devm_platform_ioremap_resource() helper for cleanner code and easier resource management. Signed-off-by:
-
Tudor Ambarus authored
Use the resource-managed kzalloc to simplify error logic. Memory allocated with this function is automatically freed on driver detach. Use struct_size() helper to calculate the size of the atdma structure with its trailing flexible array. While here, move the mem allocation higher in the probe method, as failing to allocate memory indicates a serious system issue, and everything else does not matter anyway. All these help the code look a bit cleaner. Signed-off-by:
-
Tudor Ambarus authored
Introduce a method to get the residue for a hardware linked list transfer. It makes the code easier to read. Signed-off-by:
-
Tudor Ambarus authored
Use dmaengine terminology and rename the method to better indicate what it does: it gets the residue value which will be later on set with dma_set_residue(). Signed-off-by:
-
Tudor Ambarus authored
struct dma_tx_state defines residue as u32. atc_get_bytes_left() returned an int which could be either an error or the value of the residue. This could cause problems if the controller supported a u32 buffer transfer size and the u32 value was past the max int can hold. Our controller does not support u32 buffer transfer size, but even so, improve the code and pass the residue by address to avoid unnecessary implicit casts and make atc_get_bytes_left() return 0 on success or -errno on errors. Signed-off-by:
-
Tudor Ambarus authored
Conversions of void * are applied automatically when other pointer types are assigned to and from void *. Remove the superfluous cast. Signed-off-by:
-
Tudor Ambarus authored
txstate is an optional parameter used to get a struct with auxilary transfer status information. When not provided the call to device_tx_status() should return the status of the dma cookie. Return the status of dma cookie when the txstate optional parameter is not provided. Signed-off-by:
-
Tudor Ambarus authored
The descriptor was already removed from the transfer list, there's no reason to keep the channel lock while printing desc info, thus do the prints without holding the lock. Signed-off-by:
-
Vinod Koul authored
Merge due to at_hdmac driver dependency
-
- 08 Nov, 2022 18 commits
-
-
Tudor Ambarus authored
dma_async_device_register() can fail, check the return code and display an error. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Signed-off-by:
-
Tudor Ambarus authored
The iterator can not be greater than ATC_MAX_DSCR_TRIALS, as the for loop will stop when i == ATC_MAX_DSCR_TRIALS. While here, use the common "i" name for the iterator. Fixes: 93dce3a6 ("dmaengine: at_hdmac: fix residue computation") Signed-off-by:
-
Tudor Ambarus authored
at_hdmac uses __raw_writel for register writes. In the absence of a barrier, the CPU may reorder the register operations. Introduce a write memory barrier so that the CPU does not reorder the channel enable, thus the start of the transfer, without making sure that all the pre-required register fields are already written. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
Peter Rosin <peda@axentia.se> Signed-off-by:
-
Tudor Ambarus authored
In case the controller detected an error, the code took the chance to move all the queued (submitted) descriptors to the active (issued) list. This was wrong as if there were any descriptors in the submitted list they were moved to the issued list without actually issuing them to the controller, thus a completion could be raised without even fireing the descriptor. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
As it was before, the descriptor was issued to the hardware without adding it to the active (issued) list. This could result in a completion of other descriptor, or/and in the descriptor never being completed. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
The tasklet (atc_advance_work()) did not held the channel lock when retrieving the first active descriptor, causing concurrency problems if issue_pending() was called in between. If issue_pending() was called exactly after the lock was released in the tasklet (atc_advance_work()), atc_chain_complete() could complete a descriptor for which the controller has not yet raised an interrupt. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
There's no need to hold the channel lock when freeing the memset buf, as the operation has already completed. Free the memset buf without holding the channel lock. Fixes: 4d112426 ("dmaengine: hdmac: Add memset capabilities") Signed-off-by:
-
Tudor Ambarus authored
The descriptor was added to the free_list before calling the callback, which could result in reissuing of the same descriptor and calling of a single callback for both. Move the decriptor to the free list after the callback is invoked. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
atc_complete_all() had concurrency bugs, thus remove it: 1/ atc_complete_all() in its entirety was buggy, as when the atchan->queue list (the one that contains descriptors that are not yet issued to the hardware) contained descriptors, it fired just the first from the atchan->queue, but moved all the desc from atchan->queue to atchan->active_list and considered them all as fired. This could result in calling the completion of a descriptor that was not yet issued to the hardware. 2/ when in tasklet at atc_advance_work() time, atchan->active_list was queried without holding the lock of the chan. This can result in atchan->active_list concurrency problems between the tasklet and issue_pending(). Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
Now that the complete callback call was removed from device_terminate_all(), we can protect the atchan->status with the channel lock. The atomic bitops on atchan->status do not substitute proper locking on the status, as one could still modify the status after the lock was dropped in atc_terminate_all() but before the atomic bitops were executed. Fixes: 078a6506 ("dmaengine: at_hdmac: Fix deadlocks") Reported-by:
-
Tudor Ambarus authored
The method was wrong because it violated the dmaengine API. For aborted transfers the complete callback should not be called. Fix the behavior and do not call the complete callback on device_terminate_all. Fixes: 808347f6 ("dmaengine: at_hdmac: add DMA slave transfers") Reported-by:
-
Tudor Ambarus authored
Multiple calls to atc_issue_pending() could result in a premature completion of a descriptor from the atchan->active list, as the method always completed the first active descriptor from the list. Instead, issue_pending() should just take the first transaction descriptor from the pending queue, move it to active_list and start the transfer. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
Cyclic channels must too call issue_pending in order to start a transfer. Start the transfer in issue_pending regardless of the type of channel. This wrongly worked before, because in the past the transfer was started at tx_submit level when only a desc in the transfer list. Fixes: 53830cc7 ("dmaengine: at_hdmac: add cyclic DMA operation support") Reported-by:
-
Tudor Ambarus authored
tx_submit is supposed to push the current transaction descriptor to a pending queue, waiting for issue_pending() to be called. issue_pending() must start the transfer, not tx_submit(), thus remove atc_dostart() from atc_tx_submit(). Clients of at_xdmac that assume that tx_submit() starts the transfer must be updated and call dma_async_issue_pending() if they miss to call it. The vdbg print was moved to after the lock is released. It is desirable to do the prints without the lock held if possible, and because the if statement disappears there's no reason why to do the print while holding the lock. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Reported-by:
-
Tudor Ambarus authored
Those hardware registers are all of 32 bits, while dma_addr_t ca be of type u64 or u32 depending on CONFIG_ARCH_DMA_ADDR_T_64BIT. Force u32 to comply with what the hardware expects. Fixes: dc78baa2 ("dmaengine: at_hdmac: new driver for the Atmel AHB DMA Controller") Signed-off-by:
-
Amelie Delaunay authored
When disabling dma channel, a TCF flag is set and as TCIE is enabled, an interrupt is raised. On a busy system, the interrupt may have latency and the user can ask for dmaengine_resume while stm32-dma driver has not yet managed the complete pause (backup of registers to restore state in resume). To avoid such a case, instead of waiting the interrupt to backup the registers, do it just after disabling the channel and discard Transfer Complete interrupt in case the channel is paused. Fixes: 099a9a94 ("dmaengine: stm32-dma: add device_pause/device_resume support") Signed-off-by:
Amelie Delaunay <amelie.delaunay@foss.st.com> Link: https://lore.kernel.org/r/20221024083611.132588-1-amelie.delaunay@foss.st.comSigned-off-by:
-
Yang Yingliang authored
If device_register() fails, it should call put_device() to give up reference, the name allocated in dev_set_name() can be freed in callback function kobject_cleanup(). Fixes: 5b65781d ("dmaengine: ti: k3-udma-glue: Add support for K3 PKTDMA") Signed-off-by:
Yang Yingliang <yangyingliang@huawei.com> Acked-by:
Peter Ujfalusi <peter.ujfalusi@gmail.com> Link: https://lore.kernel.org/r/20221020062827.2914148-1-yangyingliang@huawei.comSigned-off-by:
-
Christophe JAILLET authored
A clk_prepare_enable() call in the probe is not balanced by a corresponding clk_disable_unprepare() in the remove function. Add the missing call. Fixes: 3cd2c313 ("dmaengine: mv_xor_v2: Fix clock resource by adding a register clock") Signed-off-by:
Christophe JAILLET <christophe.jaillet@wanadoo.fr> Link: https://lore.kernel.org/r/e9e3837a680c9bd2438e4db2b83270c6c052d005.1666640987.git.christophe.jaillet@wanadoo.frSigned-off-by:
-
