Thanks Lucas. Let's wait for comments from Vinod. -----Original Message----- From: Lucas Stach [mailto:l.stach@xxxxxxxxxxxxxx] Sent: 2018年6月26日 23:04 To: Robin Gong <yibin.gong@xxxxxxx>; vkoul@xxxxxxxxxx; s.hauer@xxxxxxxxxxxxxx; dan.j.williams@xxxxxxxxx; gregkh@xxxxxxxxxxxxxxxxxxx; jslaby@xxxxxxxx Cc: linux-serial@xxxxxxxxxxxxxxx; dmaengine@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; dl-linux-imx <linux-imx@xxxxxxx> Subject: Re: [PATCH v5 0/7] add virt-dma support for imx-sdma Hi Robin, I've tested this whole series with the SDMA being used for SPI, UART and SSI with no regressions spotted. As this should cover most common use-cases, I think this series is good to go in. Tested-by: Lucas Stach <l.stach@xxxxxxxxxxxxxx> Regards, Lucas Am Mittwoch, den 20.06.2018, 00:56 +0800 schrieb Robin Gong: > The legacy sdma driver has below limitations or drawbacks: > 1. Hardcode the max BDs number as "PAGE_SIZE / sizeof(*)", and alloc > one page size for one channel regardless of only few BDs needed > most time. But in few cases, the max PAGE_SIZE maybe not enough. > 2. One SDMA channel can't stop immediatley once channel disabled > which > means SDMA interrupt may come in after this channel > terminated.There > are some patches for this corner case such as commit > "2746e2c389f9", > but not cover non-cyclic. > > The common virt-dma overcomes the above limitations. It can alloc bd > dynamically and free bd once this tx transfer done. No memory wasted > or maximum limititation here, only depends on how many memory can be > requested from kernel. For No.2, such issue can be workaround by > checking if there is available descript("sdmac->desc") now once the > unwanted interrupt coming. At last the common virt-dma is easier for sdma driver maintain. > > Change from v4: > 1. identify lockdep issue which caused by allocate memory with > 'GFP_KERNEL', change to 'GFP_NOWAIT' instead so that lockdep > ignore check. That also make sense since Audio/uart driver may > call dma function after spin_lock_irqsave()... > 2. use dma pool instead for bd description allocated,since audio > driver may call dma_terminate_all in irq. Please refer to 7/7. > 3. remove 7/7 serial patch in v4, since lockdep issued fixed by No.1 > > Change from v3: > 1. add two uart patches which impacted by this patchset. > 2. unlock 'vc.lock' before cyclic dma callback and lock again after > it because some driver such as uart will call dmaengine_tx_status > which will acquire 'vc.lock' again and dead lock comes out. > 3. remove 'Revert commit' stuff since that patch is not wrong and > combine two patch into one patch as Sascha's comment. > > Change from v2: > 1. include Sascha's patch to make the main patch easier to review. > Thanks Sacha. > 2. remove useless 'desc'/'chan' in struct sdma_channe. > > Change from v1: > 1. split v1 patch into 5 patches. > 2. remove some unnecessary condition check. > 3. remove unnecessary 'pending' list. > > Robin Gong (6): > tty: serial: imx: correct dma cookie status > dmaengine: imx-sdma: add virt-dma support > dmaengine: imx-sdma: remove useless 'lock' and 'enabled' in 'struct > sdma_channel' > dmaengine: imx-sdma: remove the maximum limitation for bd numbers > dmaengine: imx-sdma: add sdma_transfer_init to decrease code overlap > dmaengine: imx-sdma: alloclate bd memory from dma pool > > Sascha Hauer (1): > dmaengine: imx-sdma: factor out a struct sdma_desc from struct > sdma_channel > > drivers/dma/Kconfig | 1 + > drivers/dma/imx-sdma.c | 400 > +++++++++++++++++++++++++++-------------------- > drivers/tty/serial/imx.c | 2 +- > 3 files changed, 235 insertions(+), 168 deletions(-) > ��.n��������+%������w��{.n�����{��ǫ����{ay� ʇڙ���f���h������_�(�階�ݢj"��������G����?���&��
