Re: [PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
On Tue, Oct 15, 2019 at 04:19:55PM +0300, Dan Carpenter wrote: > On Tue, Oct 15, 2019 at 01:24:50PM +0200, Alexander Gordeev wrote: > > On Thu, Oct 10, 2019 at 02:30:34PM +0300, Dan Carpenter wrote: > > > On Thu, Oct 10, 2019 at 10:51:45AM +0200, Alexander Gordeev wrote: > > > > On Wed, Oct 09, 2019 at 09:53:23PM +0300, Dan Carpenter wrote: > > > > > > > > + u32 *rd_flags = hw->dma_desc_table_rd.cpu_addr->flags; > > > > > > > > + u32 *wr_flags = hw->dma_desc_table_wr.cpu_addr->flags; > > > > > > > > + struct avalon_dma_desc *desc; > > > > > > > > + struct virt_dma_desc *vdesc; > > > > > > > > + bool rd_done; > > > > > > > > + bool wr_done; > > > > > > > > + > > > > > > > > + spin_lock(lock); > > > > [*] > > > > > > > > > > + > > > > > > > > + rd_done = (hw->h2d_last_id < 0); > > > > > > > > + wr_done = (hw->d2h_last_id < 0); > > > > > > > > + > > > > > > > > + if (rd_done && wr_done) { > > > > > > > > + spin_unlock(lock); > > > > > > > > + return IRQ_NONE; > > > > > > > > + } > > > > > > > > + > > > > > > > > + do { > > > > > > > > + if (!rd_done && rd_flags[hw->h2d_last_id]) > > > > > > > > + rd_done = true; > > > > > > > > + > > > > > > > > + if (!wr_done && wr_flags[hw->d2h_last_id]) > > > > > > > > + wr_done = true; > > > > > > > > + } while (!rd_done || !wr_done); > > > Thinking about this some more, my initial instinct was still correct > actually. If we're holding the lock to prevent the CPU from writing > to it then how is hw->d2h_last_id updated in the other thread? Either > it must deadlock or it must be a race condition. hw->d2h_last_id and hw->h2d_last_id indexes are not expected to get updated while within the handler. It is wr_flags[hw->d2h_last_id] and/or rd_flags[hw->h2d_last_id] that should be set from zero to one by the DMA controller once/before it sends the MSI interrupt. Therefore, once we're in the handler, either wr_flags[hw->d2h_last_id] or rd_flags[hw->h2d_last_id] should be non- zero. However, the Intel documentation uses a suspicious wording for description of the above: "The Descriptor Controller writes a 1 to the done bit of the status DWORD to indicate successful completion. The Descriptor Controller also sends an MSI interrupt for the final descriptor. After receiving this MSI, host software can poll the done bit to determine status." (The "the status DWORD" in the excerpt above are located in coherent DMA memory-mapped arrays wr_flags[hw->d2h_last_id] and rd_flags[hw->h2d_last_id]) The confusing statement is "After receiving this MSI, host software can poll the done bit to determine status." Why would one poll a bit *after* the MSI received? In good design it should be set by the time the MSI arrived. May be they meant "checked" rather than "polled" or implied the CPU still could see zero in the status DWORD due to DMA memory incoherency.. Anyways, that has never been observed and I added the loop out of the paranoia, it never loops for me. HTH > regards, > dan carpenter > ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
Re: [PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
On Tue, Oct 15, 2019 at 04:03:21PM +0530, Vinod Koul wrote: > what kind of device is this? I dont think we want these and the ones > coming below as part of kernel kconfig! Yes, I have already been pointed out on this and will put those as kernel module parameters in the next version. The device is an IP block for Intel FPGAs. > > +static int start_dma_xfer(struct avalon_dma_hw *hw, > > + struct avalon_dma_desc *desc) > > +{ > > + size_t ctrl_off; > > + struct __dma_desc_table *__table; > > + struct dma_desc_table *table; > > + u32 rc_src_hi, rc_src_lo; > > + u32 ep_dst_lo, ep_dst_hi; > > + int last_id, *__last_id; > > + int nr_descs; > > + > > + if (desc->direction == DMA_TO_DEVICE) { > > + __table = >dma_desc_table_rd; > > + > > + ctrl_off = AVALON_DMA_RD_CTRL_OFFSET; > > + > > + ep_dst_hi = AVALON_DMA_RD_EP_DST_HI; > > + ep_dst_lo = AVALON_DMA_RD_EP_DST_LO; > > + > > + __last_id = >h2d_last_id; > > + } else if (desc->direction == DMA_FROM_DEVICE) { > > + __table = >dma_desc_table_wr; > > + > > + ctrl_off = AVALON_DMA_WR_CTRL_OFFSET; > > + > > + ep_dst_hi = AVALON_DMA_WR_EP_DST_HI; > > + ep_dst_lo = AVALON_DMA_WR_EP_DST_LO; > > + > > + __last_id = >d2h_last_id; > > + } else { > > + BUG(); > > + } > > + > > + table = __table->cpu_addr; > > + memset(>flags, 0, sizeof(table->flags)); > > + > > + nr_descs = setup_dma_descs(table->descs, desc); > > + if (WARN_ON(nr_descs < 1)) > > + return nr_descs; (1) Here it may fail. > > + last_id = nr_descs - 1; > > + *__last_id = last_id; > > + > > + rc_src_hi = __table->dma_addr >> 32; > > + rc_src_lo = (u32)__table->dma_addr; > > + > > + start_xfer(hw->regs, ctrl_off, > > + rc_src_hi, rc_src_lo, > > + ep_dst_hi, ep_dst_lo, > > + last_id); > > + > > + return 0; > > why have a return when you always return 0? Please see (1) above. > > +static irqreturn_t avalon_dma_interrupt(int irq, void *dev_id) > > +{ > > + struct avalon_dma *adma = (struct avalon_dma *)dev_id; > > + struct avalon_dma_chan *chan = >chan; > > + struct avalon_dma_hw *hw = >hw; > > + spinlock_t *lock = >vchan.lock; > > + u32 *rd_flags = hw->dma_desc_table_rd.cpu_addr->flags; > > + u32 *wr_flags = hw->dma_desc_table_wr.cpu_addr->flags; (2) > > + struct avalon_dma_desc *desc; > > + struct virt_dma_desc *vdesc; > > + bool rd_done; > > + bool wr_done; > > + > > + spin_lock(lock); > > + > > + rd_done = (hw->h2d_last_id < 0); > > + wr_done = (hw->d2h_last_id < 0); (3) > > + > > + if (rd_done && wr_done) { > > + spin_unlock(lock); > > + return IRQ_NONE; > > + } > > + > > + do { > > + if (!rd_done && rd_flags[hw->h2d_last_id]) > > + rd_done = true; > > + > > + if (!wr_done && wr_flags[hw->d2h_last_id]) > > + wr_done = true; > > + } while (!rd_done || !wr_done); > > Can you explain this logic. I dont like busy loop in isr and who updates Here is the comment in the next version of the patch: The Intel documentation claims "The Descriptor Controller writes a 1 to the done bit of the status DWORD to indicate successful completion. The Descriptor Controller also sends an MSI interrupt for the final descriptor. After receiving this MSI, host software can poll the done bit to determine status." The above could be read like MSI interrupt might be delivered before the corresponding done bit is set. But in reality it does not happen at all (or happens really rare). So put here the done bit polling, just in case. > rd_done and rd_flags etc..? rd_flags/wr_flags are updated by the hardware (2) and mapped in as coherent DMA memory. rd_done/wr_done are just local variables (3) > > +static int __init avalon_pci_probe(struct pci_dev *pci_dev, > > + const struct pci_device_id *id) > > +{ > > + void *adma; > > + void __iomem *regs; > > + int ret; > > + > > + ret = pci_enable_device(pci_dev); > > + if (ret) > > + goto enable_err; > > + > > + ret = pci_request_regions(pci_dev, DRIVER_NAME); > > + if (ret) > > + goto reg_err; > > + > > + regs = pci_ioremap_bar(pci_dev, PCI_BAR); > > This is a pci device, so we should really be using the PCI way of > setting and querying the resources. Doing this thru kernel config > options is not correct! I assume you are referring to PCI_BAR in this particular case, right? If so, this is an excerpt from the documentation that I read like a BAR the DMA controller is mapped to needs to be configurable: "When the DMA Descriptor Controller is externally instantiated, these registers are accessed through a BAR. The offsets must be added to the base address for the read controller. When the Descriptor Controller is internally instantiated these registers are accessed through BAR0." Thank
Re: [PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
On Tue, Oct 15, 2019 at 02:41:08PM +0300, Dan Carpenter wrote: > > > > > > > > + spin_lock(lock); > > > > [*] > > [ snip ] > > > I struggle to realize how the spinlock I use (see [*] above) does not > > protect the reader. > > Argh I'm really sorry. I completely didn't see the spinlock. :P Np ;) May be in the next version it will be more visible. I done it the way you asked ( even though I do not like it :D ): spin_lock(>vchan.lock); > I am embarrassed. Wow... > > regards, > dan carpenter > ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
Re: [PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
On Thu, Oct 10, 2019 at 02:30:34PM +0300, Dan Carpenter wrote: > On Thu, Oct 10, 2019 at 10:51:45AM +0200, Alexander Gordeev wrote: > > On Wed, Oct 09, 2019 at 09:53:23PM +0300, Dan Carpenter wrote: > > > > > > + u32 *rd_flags = hw->dma_desc_table_rd.cpu_addr->flags; > > > > > > + u32 *wr_flags = hw->dma_desc_table_wr.cpu_addr->flags; > > > > > > + struct avalon_dma_desc *desc; > > > > > > + struct virt_dma_desc *vdesc; > > > > > > + bool rd_done; > > > > > > + bool wr_done; > > > > > > + > > > > > > + spin_lock(lock); [*] > > > > > > + > > > > > > + rd_done = (hw->h2d_last_id < 0); > > > > > > + wr_done = (hw->d2h_last_id < 0); > > > > > > + > > > > > > + if (rd_done && wr_done) { > > > > > > + spin_unlock(lock); > > > > > > + return IRQ_NONE; > > > > > > + } > > > > > > + > > > > > > + do { > > > > > > + if (!rd_done && rd_flags[hw->h2d_last_id]) > > > > > > + rd_done = true; > > > > > > + > > > > > > + if (!wr_done && wr_flags[hw->d2h_last_id]) > > > > > > + wr_done = true; > > > > > > + } while (!rd_done || !wr_done); > > > > > > > > > > This loop is very strange. It feels like the last_id indexes needs > > > > > to atomic or protected from racing somehow so we don't do an out of > > > > > bounds read. > > > > [...] > > > > > You're missing my point. When we set > > > hw->d2h_last_id = 1; > > [1] > > > ... > > > hw->d2h_last_id = 2; > > [2] > > > > > There is a tiny moment where ->d2h_last_id is transitioning from 1 to 2 > > > where its value is unknown. We're in a busy loop here so we have a > > > decent chance of hitting that 1/1000,000th of a second. If we happen to > > > hit it at exactly the right time then we're reading from a random > > > address and it will cause an oops. > > > > > > We have to use atomic_t types or something to handle race conditions. > > > > Err.. I am still missing the point :( In your example I do see a chance > > for a reader to read out 1 at point in time [2] - because of SMP race. > > But what could it be other than 1 or 2? > > > > The 1 to 2 transition was a poorly chosen example, but a -1 to 1 > trasition is better. The cpu could write a byte at a time. So maybe > it only wrote the two highest bytes so now it's 0x. It's not -1 and > it's not 1 and it's not a valid index. > > > Anyways, all code paths dealing with h2d_last_id and d2h_last_id indexes > > are protected with a spinlock. > > You have to protect both the writer and the reader. (That's why this > bug is so easy to spot). https://lwn.net/Articles/793253/ I struggle to realize how the spinlock I use (see [*] above) does not protect the reader. I am going to post updated version shortly, hopefully it will make more sense. > regards, > dan carpenter > ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
Re: [PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
On Wed, Oct 09, 2019 at 09:53:23PM +0300, Dan Carpenter wrote: > > > > + u32 *rd_flags = hw->dma_desc_table_rd.cpu_addr->flags; > > > > + u32 *wr_flags = hw->dma_desc_table_wr.cpu_addr->flags; > > > > + struct avalon_dma_desc *desc; > > > > + struct virt_dma_desc *vdesc; > > > > + bool rd_done; > > > > + bool wr_done; > > > > + > > > > + spin_lock(lock); > > > > + > > > > + rd_done = (hw->h2d_last_id < 0); > > > > + wr_done = (hw->d2h_last_id < 0); > > > > + > > > > + if (rd_done && wr_done) { > > > > + spin_unlock(lock); > > > > + return IRQ_NONE; > > > > + } > > > > + > > > > + do { > > > > + if (!rd_done && rd_flags[hw->h2d_last_id]) > > > > + rd_done = true; > > > > + > > > > + if (!wr_done && wr_flags[hw->d2h_last_id]) > > > > + wr_done = true; > > > > + } while (!rd_done || !wr_done); > > > > > > This loop is very strange. It feels like the last_id indexes needs > > > to atomic or protected from racing somehow so we don't do an out of > > > bounds read. [...] > You're missing my point. When we set > hw->d2h_last_id = 1; [1] > ... > hw->d2h_last_id = 2; [2] > There is a tiny moment where ->d2h_last_id is transitioning from 1 to 2 > where its value is unknown. We're in a busy loop here so we have a > decent chance of hitting that 1/1000,000th of a second. If we happen to > hit it at exactly the right time then we're reading from a random > address and it will cause an oops. > > We have to use atomic_t types or something to handle race conditions. Err.. I am still missing the point :( In your example I do see a chance for a reader to read out 1 at point in time [2] - because of SMP race. But what could it be other than 1 or 2? Anyways, all code paths dealing with h2d_last_id and d2h_last_id indexes are protected with a spinlock. > regards, > dan carpenter > ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
Re: [PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
On Wed, Oct 09, 2019 at 03:14:41PM +0300, Dan Carpenter wrote: > > +config AVALON_DMA_PCI_VENDOR_ID > > + hex "PCI vendor ID" > > + default "0x1172" > > + > > +config AVALON_DMA_PCI_DEVICE_ID > > + hex "PCI device ID" > > + default "0xe003" > > This feels wrong. Why isn't it known in advance. Because device designers would likely use they own IDs. The ones I put are just defaults inherited from the (Altera) reference design. > > + u32 *rd_flags = hw->dma_desc_table_rd.cpu_addr->flags; > > + u32 *wr_flags = hw->dma_desc_table_wr.cpu_addr->flags; > > + struct avalon_dma_desc *desc; > > + struct virt_dma_desc *vdesc; > > + bool rd_done; > > + bool wr_done; > > + > > + spin_lock(lock); > > + > > + rd_done = (hw->h2d_last_id < 0); > > + wr_done = (hw->d2h_last_id < 0); > > + > > + if (rd_done && wr_done) { > > + spin_unlock(lock); > > + return IRQ_NONE; > > + } > > + > > + do { > > + if (!rd_done && rd_flags[hw->h2d_last_id]) > > + rd_done = true; > > + > > + if (!wr_done && wr_flags[hw->d2h_last_id]) > > + wr_done = true; > > + } while (!rd_done || !wr_done); > > This loop is very strange. It feels like the last_id indexes needs > to atomic or protected from racing somehow so we don't do an out of > bounds read. My bad. I should have put a comment on this. This polling comes from my reading of the Intel documentation: "The MSI interrupt notifies the host when a DMA operation has completed. After the host receives this interrupt, it can poll the DMA read or write status table to determine which entry or entries have the done bit set." "The Descriptor Controller writes a 1 to the done bit of the status DWORD to indicate successful completion. The Descriptor Controller also sends an MSI interrupt for the final descriptor. After receiving this MSI, host software can poll the done bit to determine status." I sense an ambiguity above. It sounds possible an MSI interrupt could be delivered before corresponding done bit is set. May be imperfect wording.. Anyway, the loop does look weird and in reality I doubt I observed the done bit unset even once. So I put this polling just in case. > > + struct avalon_dma_chan *chan = to_avalon_dma_chan(dma_chan); > > + struct avalon_dma_desc *desc; > > + gfp_t gfp_flags = in_interrupt() ? GFP_NOWAIT : GFP_KERNEL; > > + dma_addr_t dev_addr; > > + > > + if (direction == DMA_MEM_TO_DEV) > > + dev_addr = chan->dst_addr; > > + else if (direction == DMA_DEV_TO_MEM) > > + dev_addr = chan->src_addr; > > + else > > + return NULL; > > + > > + desc = kzalloc(sizeof(*desc), gfp_flags); > > Everyone else does GFP_WAIT or GFP_ATOMIC. Is GFP_KERNEL really okay? I am not sure why not to use GFP_KERNEL from non-atomic context. Documentation/driver-api/dmaengine/provider.rst claims always to use GFP_NOWAIT though: - Any allocation you might do should be using the GFP_NOWAIT flag, in order not to potentially sleep, but without depleting the emergency pool either. So probably I just should use GFP_NOWAIT. Thanks, Dan! > regards, > dan carpenter > ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
Re: [PATCH RFC 0/2] staging: Support Avalon-MM DMA Interface for PCIe
On Thu, Sep 19, 2019 at 01:37:08PM +0200, Greg KH wrote: > Why is this being submitted for drivers/staging/ and not the "real" part > of the kernel tree? Hi Greg! I sent v2 of the patchset, but it does not need to be part of the staging tree. I CC-ed de...@driverdev.osuosl.org for reference. Thanks! > All staging code must have a TODO file listing what needs to be done in > order to get it out of staging, and be self-contained (i.e. no files > include/linux/) > > Please fix that up when resending this series. > > thanks, > > greg k-h ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
[PATCH RFC v2 2/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe test
This is sample implementation of a driver that uses "avalon-dma" driver interface to perform data transfers between on-chip and system memory in devices using Avalon-MM DMA Interface for PCIe design. Companion user-level tool could be found at https://github.com/a-gordeev/avalon-tool.git CC: Michael Chen CC: de...@driverdev.osuosl.org CC: dmaeng...@vger.kernel.org Signed-off-by: Alexander Gordeev --- drivers/dma/Kconfig | 1 + drivers/dma/Makefile| 1 + drivers/dma/avalon-test/Kconfig | 23 + drivers/dma/avalon-test/Makefile| 14 + drivers/dma/avalon-test/avalon-dev.c| 65 +++ drivers/dma/avalon-test/avalon-dev.h| 33 ++ drivers/dma/avalon-test/avalon-ioctl.c | 128 + drivers/dma/avalon-test/avalon-ioctl.h | 12 + drivers/dma/avalon-test/avalon-mmap.c | 93 drivers/dma/avalon-test/avalon-mmap.h | 12 + drivers/dma/avalon-test/avalon-sg-buf.c | 132 + drivers/dma/avalon-test/avalon-sg-buf.h | 26 + drivers/dma/avalon-test/avalon-util.c | 54 ++ drivers/dma/avalon-test/avalon-util.h | 12 + drivers/dma/avalon-test/avalon-xfer.c | 697 drivers/dma/avalon-test/avalon-xfer.h | 28 + include/uapi/linux/avalon-ioctl.h | 32 ++ 17 files changed, 1363 insertions(+) create mode 100644 drivers/dma/avalon-test/Kconfig create mode 100644 drivers/dma/avalon-test/Makefile create mode 100644 drivers/dma/avalon-test/avalon-dev.c create mode 100644 drivers/dma/avalon-test/avalon-dev.h create mode 100644 drivers/dma/avalon-test/avalon-ioctl.c create mode 100644 drivers/dma/avalon-test/avalon-ioctl.h create mode 100644 drivers/dma/avalon-test/avalon-mmap.c create mode 100644 drivers/dma/avalon-test/avalon-mmap.h create mode 100644 drivers/dma/avalon-test/avalon-sg-buf.c create mode 100644 drivers/dma/avalon-test/avalon-sg-buf.h create mode 100644 drivers/dma/avalon-test/avalon-util.c create mode 100644 drivers/dma/avalon-test/avalon-util.h create mode 100644 drivers/dma/avalon-test/avalon-xfer.c create mode 100644 drivers/dma/avalon-test/avalon-xfer.h create mode 100644 include/uapi/linux/avalon-ioctl.h diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig index f6f43480a4a4..4b3c6a6baf4c 100644 --- a/drivers/dma/Kconfig +++ b/drivers/dma/Kconfig @@ -670,6 +670,7 @@ source "drivers/dma/sh/Kconfig" source "drivers/dma/ti/Kconfig" source "drivers/dma/avalon/Kconfig" +source "drivers/dma/avalon-test/Kconfig" # clients comment "DMA Clients" diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile index fd7e11417b73..eb3ee7f6cac6 100644 --- a/drivers/dma/Makefile +++ b/drivers/dma/Makefile @@ -76,6 +76,7 @@ obj-$(CONFIG_XGENE_DMA) += xgene-dma.o obj-$(CONFIG_ZX_DMA) += zx_dma.o obj-$(CONFIG_ST_FDMA) += st_fdma.o obj-$(CONFIG_AVALON_DMA) += avalon/ +obj-$(CONFIG_AVALON_TEST) += avalon-test/ obj-y += mediatek/ obj-y += qcom/ diff --git a/drivers/dma/avalon-test/Kconfig b/drivers/dma/avalon-test/Kconfig new file mode 100644 index ..021c28fe77a6 --- /dev/null +++ b/drivers/dma/avalon-test/Kconfig @@ -0,0 +1,23 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA engine +# +# Author: Alexander Gordeev +# +config AVALON_TEST + select AVALON_DMA + tristate "Intel Avalon-MM DMA Interface for PCIe test driver" + help + This selects a test driver for Avalon-MM DMA Interface for PCI + +if AVALON_TEST + +config AVALON_TEST_TARGET_BASE + hex "Target device base address" + default "0x7000" + +config AVALON_TEST_TARGET_SIZE + hex "Target device memory size" + default "0x1000" + +endif diff --git a/drivers/dma/avalon-test/Makefile b/drivers/dma/avalon-test/Makefile new file mode 100644 index ..63351c52478a --- /dev/null +++ b/drivers/dma/avalon-test/Makefile @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA driver +# +# Author: Alexander Gordeev +# +obj-$(CONFIG_AVALON_TEST) += avalon-test.o + +avalon-test-objs :=avalon-dev.o \ + avalon-ioctl.o \ + avalon-mmap.o \ + avalon-sg-buf.o \ + avalon-xfer.o \ + avalon-util.o diff --git a/drivers/dma/avalon-test/avalon-dev.c b/drivers/dma/avalon-test/avalon-dev.c new file mode 100644 index ..9e83f777f937 --- /dev/null +++ b/drivers/dma/avalon-test/avalon-dev.c @@ -0,0 +1,65 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Avalon DMA driver + * + * Author: Alexander Gordeev + */ +#include +#include +#include + +#include "avalon-dev.h" +#include "avalon-ioctl.h" +#include "avalon-mmap.h" + +const struct file_operations avalon_dev_fops = { + .llseek = generic_file_llseek, + .unlocked_ioctl = avalon_dev_ioctl, + .mmap = avalo
[PATCH v2 1/2] dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe
Support Avalon-MM DMA Interface for PCIe used in hard IPs for Intel Arria, Cyclone or Stratix FPGAs. CC: Michael Chen CC: de...@driverdev.osuosl.org CC: dmaeng...@vger.kernel.org Signed-off-by: Alexander Gordeev --- drivers/dma/Kconfig | 2 + drivers/dma/Makefile | 1 + drivers/dma/avalon/Kconfig | 88 +++ drivers/dma/avalon/Makefile | 6 + drivers/dma/avalon/avalon-core.c | 432 +++ drivers/dma/avalon/avalon-core.h | 90 +++ drivers/dma/avalon/avalon-hw.c | 212 +++ drivers/dma/avalon/avalon-hw.h | 86 ++ drivers/dma/avalon/avalon-pci.c | 150 +++ 9 files changed, 1067 insertions(+) create mode 100644 drivers/dma/avalon/Kconfig create mode 100644 drivers/dma/avalon/Makefile create mode 100644 drivers/dma/avalon/avalon-core.c create mode 100644 drivers/dma/avalon/avalon-core.h create mode 100644 drivers/dma/avalon/avalon-hw.c create mode 100644 drivers/dma/avalon/avalon-hw.h create mode 100644 drivers/dma/avalon/avalon-pci.c diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig index 7af874b69ffb..f6f43480a4a4 100644 --- a/drivers/dma/Kconfig +++ b/drivers/dma/Kconfig @@ -669,6 +669,8 @@ source "drivers/dma/sh/Kconfig" source "drivers/dma/ti/Kconfig" +source "drivers/dma/avalon/Kconfig" + # clients comment "DMA Clients" depends on DMA_ENGINE diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile index f5ce8665e944..fd7e11417b73 100644 --- a/drivers/dma/Makefile +++ b/drivers/dma/Makefile @@ -75,6 +75,7 @@ obj-$(CONFIG_UNIPHIER_MDMAC) += uniphier-mdmac.o obj-$(CONFIG_XGENE_DMA) += xgene-dma.o obj-$(CONFIG_ZX_DMA) += zx_dma.o obj-$(CONFIG_ST_FDMA) += st_fdma.o +obj-$(CONFIG_AVALON_DMA) += avalon/ obj-y += mediatek/ obj-y += qcom/ diff --git a/drivers/dma/avalon/Kconfig b/drivers/dma/avalon/Kconfig new file mode 100644 index ..09e0773fcdb2 --- /dev/null +++ b/drivers/dma/avalon/Kconfig @@ -0,0 +1,88 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA engine +# +# Author: Alexander Gordeev +# +config AVALON_DMA + tristate "Intel Avalon-MM DMA Interface for PCIe" + depends on PCI + select DMA_ENGINE + select DMA_VIRTUAL_CHANNELS + help + This selects a driver for Avalon-MM DMA Interface for PCIe + hard IP block used in Intel Arria, Cyclone or Stratix FPGAs. + +if AVALON_DMA + +config AVALON_DMA_MASK_WIDTH + int "Avalon DMA streaming and coherent bitmask width" + range 0 64 + default 64 + help + Width of bitmask for streaming and coherent DMA operations + +config AVALON_DMA_CTRL_BASE + hex "Avalon DMA controllers base" + default "0x" + +config AVALON_DMA_RD_EP_DST_LO + hex "Avalon DMA read controller base low" + default "0x8000" + help + Specifies the lower 32-bits of the base address of the read + status and descriptor table in the Root Complex memory. + +config AVALON_DMA_RD_EP_DST_HI + hex "Avalon DMA read controller base high" + default "0x" + help + Specifies the upper 32-bits of the base address of the read + status and descriptor table in the Root Complex memory. + +config AVALON_DMA_WR_EP_DST_LO + hex "Avalon DMA write controller base low" + default "0x80002000" + help + Specifies the lower 32-bits of the base address of the write + status and descriptor table in the Root Complex memory. + +config AVALON_DMA_WR_EP_DST_HI + hex "Avalon DMA write controller base high" + default "0x" + help + Specifies the upper 32-bits of the base address of the write + status and descriptor table in the Root Complex memory. + +config AVALON_DMA_PCI_VENDOR_ID + hex "PCI vendor ID" + default "0x1172" + +config AVALON_DMA_PCI_DEVICE_ID + hex "PCI device ID" + default "0xe003" + +config AVALON_DMA_PCI_BAR + int "PCI device BAR the Avalon DMA controller is mapped to" + range 0 5 + default 0 + help + Number of PCI BAR the DMA controller is mapped to + +config AVALON_DMA_PCI_MSI_COUNT_ORDER + int "Count of MSIs the PCI device provides (order)" + range 0 5 + default 5 + help + Number of vectors the PCI device uses in multiple MSI mode. + This number is provided as the power of two. + +config AVALON_DMA_PCI_MSI_VECTOR + int "Vector number the DMA controller is mapped to" + range 0 31 + default 0 + help + Number of MSI vector the DMA controller is mapped to in + multiple MSI mode. + +endif diff --git a/drivers/dma/avalon/Makefile b/drivers/dma/avalon
[PATCH v2 0/2] dmaengine: avalon: Support Avalon-MM DMA Interface for PCIe
This series is against v5.4-rc2 Changes since v1: - "avalon-dma" converted to "dmaengine" model; - "avalon-drv" renamed to "avalon-test"; The Avalon-MM DMA Interface for PCIe is a design used in hard IPs for Intel Arria, Cyclone or Stratix FPGAs. It transfers data between on-chip memory and system memory. Patch 1. This patch introduces "avalon-dma" driver that conforms to "dmaengine" model. Patch 2. The existing "dmatest" is not meant for DMA_SLAVE type of transfers needed by "avalon-dma". Instead, custom "avalon-test" driver was used to debug and stress "avalon-dma". If it could be useful for a wider audience, I can make it optional part of "avalon-dma" sources or leave it as separate driver. Marking patch 2 as RFC for now. Testing was done using a custom FPGA build with Arria 10 FPGA streaming data to target device RAM: +--++--++--++--+ | Nios CPU |<-->| RAM|<-->| Avalon |<-PCIe->| Host CPU | +--++--++--++--+ The RAM was examined for data integrity by examining RAM contents from host CPU (indirectly - checking data DMAed to the system) and from Nios CPU that has direct access to the device RAM. A companion tool using "avalon-test" driver was used to DMA files to the device: https://github.com/a-gordeev/avalon-tool.git CC: Michael Chen CC: de...@driverdev.osuosl.org CC: dmaeng...@vger.kernel.org Alexander Gordeev (2): dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe dmaengine: avalon: Intel Avalon-MM DMA Interface for PCIe test drivers/dma/Kconfig | 3 + drivers/dma/Makefile| 2 + drivers/dma/avalon-test/Kconfig | 23 + drivers/dma/avalon-test/Makefile| 14 + drivers/dma/avalon-test/avalon-dev.c| 65 +++ drivers/dma/avalon-test/avalon-dev.h| 33 ++ drivers/dma/avalon-test/avalon-ioctl.c | 128 + drivers/dma/avalon-test/avalon-ioctl.h | 12 + drivers/dma/avalon-test/avalon-mmap.c | 93 drivers/dma/avalon-test/avalon-mmap.h | 12 + drivers/dma/avalon-test/avalon-sg-buf.c | 132 + drivers/dma/avalon-test/avalon-sg-buf.h | 26 + drivers/dma/avalon-test/avalon-util.c | 54 ++ drivers/dma/avalon-test/avalon-util.h | 12 + drivers/dma/avalon-test/avalon-xfer.c | 697 drivers/dma/avalon-test/avalon-xfer.h | 28 + drivers/dma/avalon/Kconfig | 88 +++ drivers/dma/avalon/Makefile | 6 + drivers/dma/avalon/avalon-core.c| 432 +++ drivers/dma/avalon/avalon-core.h| 90 +++ drivers/dma/avalon/avalon-hw.c | 212 +++ drivers/dma/avalon/avalon-hw.h | 86 +++ drivers/dma/avalon/avalon-pci.c | 150 + include/uapi/linux/avalon-ioctl.h | 32 ++ 24 files changed, 2430 insertions(+) create mode 100644 drivers/dma/avalon-test/Kconfig create mode 100644 drivers/dma/avalon-test/Makefile create mode 100644 drivers/dma/avalon-test/avalon-dev.c create mode 100644 drivers/dma/avalon-test/avalon-dev.h create mode 100644 drivers/dma/avalon-test/avalon-ioctl.c create mode 100644 drivers/dma/avalon-test/avalon-ioctl.h create mode 100644 drivers/dma/avalon-test/avalon-mmap.c create mode 100644 drivers/dma/avalon-test/avalon-mmap.h create mode 100644 drivers/dma/avalon-test/avalon-sg-buf.c create mode 100644 drivers/dma/avalon-test/avalon-sg-buf.h create mode 100644 drivers/dma/avalon-test/avalon-util.c create mode 100644 drivers/dma/avalon-test/avalon-util.h create mode 100644 drivers/dma/avalon-test/avalon-xfer.c create mode 100644 drivers/dma/avalon-test/avalon-xfer.h create mode 100644 drivers/dma/avalon/Kconfig create mode 100644 drivers/dma/avalon/Makefile create mode 100644 drivers/dma/avalon/avalon-core.c create mode 100644 drivers/dma/avalon/avalon-core.h create mode 100644 drivers/dma/avalon/avalon-hw.c create mode 100644 drivers/dma/avalon/avalon-hw.h create mode 100644 drivers/dma/avalon/avalon-pci.c create mode 100644 include/uapi/linux/avalon-ioctl.h -- 2.23.0 ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
[PATCH RFC 1/2] staging: avalon-dma: Avalon DMA engine
Basic support for Avalon-MM DMA Interface for PCIe found in hard IPs for Intel Arria, Cyclone or Stratix FPGAs. This is an alternative (though minimal functionality) implementation of reference design driver from Intel. Unlike the reference design, the introduced interface allows submitting contiguous buffers, scatterlists and DMA completion callbacks - much like "dmaengine" does. CC: Michael Chen CC: de...@driverdev.osuosl.org CC: dmaeng...@vger.kernel.org Signed-off-by: Alexander Gordeev --- drivers/staging/Kconfig | 2 + drivers/staging/Makefile | 1 + drivers/staging/avalon-dma/Kconfig| 45 ++ drivers/staging/avalon-dma/Makefile | 11 + drivers/staging/avalon-dma/avalon-dma-core.c | 515 ++ drivers/staging/avalon-dma/avalon-dma-core.h | 52 ++ .../staging/avalon-dma/avalon-dma-interrupt.c | 118 .../staging/avalon-dma/avalon-dma-interrupt.h | 13 + drivers/staging/avalon-dma/avalon-dma-util.c | 196 +++ drivers/staging/avalon-dma/avalon-dma-util.h | 25 + include/linux/avalon-dma-hw.h | 72 +++ include/linux/avalon-dma.h| 68 +++ 12 files changed, 1118 insertions(+) create mode 100644 drivers/staging/avalon-dma/Kconfig create mode 100644 drivers/staging/avalon-dma/Makefile create mode 100644 drivers/staging/avalon-dma/avalon-dma-core.c create mode 100644 drivers/staging/avalon-dma/avalon-dma-core.h create mode 100644 drivers/staging/avalon-dma/avalon-dma-interrupt.c create mode 100644 drivers/staging/avalon-dma/avalon-dma-interrupt.h create mode 100644 drivers/staging/avalon-dma/avalon-dma-util.c create mode 100644 drivers/staging/avalon-dma/avalon-dma-util.h create mode 100644 include/linux/avalon-dma-hw.h create mode 100644 include/linux/avalon-dma.h diff --git a/drivers/staging/Kconfig b/drivers/staging/Kconfig index 7c96a01eef6c..31c732ececd1 100644 --- a/drivers/staging/Kconfig +++ b/drivers/staging/Kconfig @@ -120,4 +120,6 @@ source "drivers/staging/kpc2000/Kconfig" source "drivers/staging/isdn/Kconfig" +source "drivers/staging/avalon-dma/Kconfig" + endif # STAGING diff --git a/drivers/staging/Makefile b/drivers/staging/Makefile index fcaac9693b83..eb974cac85d3 100644 --- a/drivers/staging/Makefile +++ b/drivers/staging/Makefile @@ -50,3 +50,4 @@ obj-$(CONFIG_EROFS_FS)+= erofs/ obj-$(CONFIG_FIELDBUS_DEV) += fieldbus/ obj-$(CONFIG_KPC2000) += kpc2000/ obj-$(CONFIG_ISDN_CAPI)+= isdn/ +obj-$(CONFIG_AVALON_DMA) += avalon-dma/ diff --git a/drivers/staging/avalon-dma/Kconfig b/drivers/staging/avalon-dma/Kconfig new file mode 100644 index ..5164e990a62b --- /dev/null +++ b/drivers/staging/avalon-dma/Kconfig @@ -0,0 +1,45 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA engine +# +# Author: Alexander Gordeev +# +config AVALON_DMA + tristate "Avalon DMA engine" + help + This selects a driver for Avalon DMA engine IP block + +config AVALON_DMA_TARGET_BASE + hex "Target device base address" + default "0x7000" + depends on AVALON_DMA + +config AVALON_DMA_TARGET_SIZE + hex "Target device memory size" + default "0x1000" + depends on AVALON_DMA + +config AVALON_DMA_CTRL_BASE + hex "Avalon DMA controllers base" + default "0x" + depends on AVALON_DMA + +config AVALON_DMA_RD_EP_DST_LO + hex "Avalon DMA read controller base low" + default "0x8000" + depends on AVALON_DMA + +config AVALON_DMA_RD_EP_DST_HI + hex "Avalon DMA read controller base high" + default "0x" + depends on AVALON_DMA + +config AVALON_DMA_WR_EP_DST_LO + hex "Avalon DMA write controller base low" + default "0x80002000" + depends on AVALON_DMA + +config AVALON_DMA_WR_EP_DST_HI + hex "Avalon DMA write controller base high" + default "0x" + depends on AVALON_DMA diff --git a/drivers/staging/avalon-dma/Makefile b/drivers/staging/avalon-dma/Makefile new file mode 100644 index 0000..61cb0ee7c7a8 --- /dev/null +++ b/drivers/staging/avalon-dma/Makefile @@ -0,0 +1,11 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA engine +# +# Author: Alexander Gordeev +# +obj-$(CONFIG_AVALON_DMA) += avalon-dma.o + +avalon-dma-objs := avalon-dma-core.o \ + avalon-dma-util.o \ + avalon-dma-interrupt.o diff --git a/drivers/staging/avalon-dma/avalon-dma-core.c b/drivers/staging/avalon-dma/avalon-dma-core.c new file mode 100644 index 0000..9e90c694785f --- /dev/null +++ b/drivers/staging/avalon-dma/avalon-dma-core.c @@ -0,0 +1,515 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Avalon D
[PATCH RFC 0/2] staging: Support Avalon-MM DMA Interface for PCIe
The Avalon-MM DMA Interface for PCIe is a design found in hard IPs for Intel Arria, Cyclone or Stratix FPGAs. It transfers data between on-chip memory and system memory. This RFC is an attempt to provide a generic API: typedef void (*avalon_dma_xfer_callback)(void *dma_async_param); int avalon_dma_submit_xfer( struct avalon_dma *avalon_dma, enum dma_data_direction direction, dma_addr_t dev_addr, dma_addr_t host_addr, unsigned int size, avalon_dma_xfer_callback callback, void *callback_param); int avalon_dma_submit_xfer_sg(struct avalon_dma *avalon_dma, enum dma_data_direction direction, dma_addr_t dev_addr, struct sg_table *sg_table, avalon_dma_xfer_callback callback, void *callback_param); int avalon_dma_issue_pending(struct avalon_dma *avalon_dma); Patch 1 introduces "avalon-dma" driver that provides the above-mentioned generic interface. Patch 2 adds "avalon-drv" driver using "avalon-dma" to transfer user- provided data. This driver was used to debug and stress "avalon-dma" and could be used as a code base for other implementations. Strictly speaking, it does not need to be part of the kernel tree. A companion tool using "avalon-drv" to DMA files (not part of this patchset) is located at g...@github.com:a-gordeev/avalon-drv-tool.git The suggested interface is developed with the standard "dmaengine" in mind and could be reworked to suit it. I would appreciate, however gathering some feedback on the implemenation first - as the hardware- specific code would persist. It is also a call for testing - I only have access to a single Arria 10 device to try on. This series is against v5.3 and could be found at g...@github.com:a-gordeev/linux.git avalon-dma-engine CC: Michael Chen CC: de...@driverdev.osuosl.org CC: dmaeng...@vger.kernel.org Alexander Gordeev (2): staging: avalon-dma: Avalon DMA engine staging: avalon-drv: Avalon DMA driver drivers/staging/Kconfig | 4 + drivers/staging/Makefile | 2 + drivers/staging/avalon-dma/Kconfig| 45 ++ drivers/staging/avalon-dma/Makefile | 11 + drivers/staging/avalon-dma/avalon-dma-core.c | 515 ++ drivers/staging/avalon-dma/avalon-dma-core.h | 52 ++ .../staging/avalon-dma/avalon-dma-interrupt.c | 118 .../staging/avalon-dma/avalon-dma-interrupt.h | 13 + drivers/staging/avalon-dma/avalon-dma-util.c | 196 ++ drivers/staging/avalon-dma/avalon-dma-util.h | 25 + drivers/staging/avalon-drv/Kconfig| 34 + drivers/staging/avalon-drv/Makefile | 14 + drivers/staging/avalon-drv/avalon-drv-dev.c | 193 ++ drivers/staging/avalon-drv/avalon-drv-ioctl.c | 137 drivers/staging/avalon-drv/avalon-drv-ioctl.h | 12 + drivers/staging/avalon-drv/avalon-drv-mmap.c | 93 +++ drivers/staging/avalon-drv/avalon-drv-mmap.h | 12 + .../staging/avalon-drv/avalon-drv-sg-buf.c| 132 .../staging/avalon-drv/avalon-drv-sg-buf.h| 26 + drivers/staging/avalon-drv/avalon-drv-util.c | 54 ++ drivers/staging/avalon-drv/avalon-drv-util.h | 12 + drivers/staging/avalon-drv/avalon-drv-xfer.c | 655 ++ drivers/staging/avalon-drv/avalon-drv-xfer.h | 24 + drivers/staging/avalon-drv/avalon-drv.h | 22 + include/linux/avalon-dma-hw.h | 72 ++ include/linux/avalon-dma.h| 68 ++ include/uapi/linux/avalon-drv-ioctl.h | 30 + 27 files changed, 2571 insertions(+) create mode 100644 drivers/staging/avalon-dma/Kconfig create mode 100644 drivers/staging/avalon-dma/Makefile create mode 100644 drivers/staging/avalon-dma/avalon-dma-core.c create mode 100644 drivers/staging/avalon-dma/avalon-dma-core.h create mode 100644 drivers/staging/avalon-dma/avalon-dma-interrupt.c create mode 100644 drivers/staging/avalon-dma/avalon-dma-interrupt.h create mode 100644 drivers/staging/avalon-dma/avalon-dma-util.c create mode 100644 drivers/staging/avalon-dma/avalon-dma-util.h create mode 100644 drivers/staging/avalon-drv/Kconfig create mode 100644 drivers/staging/avalon-drv/Makefile create mode 100644 drivers/staging/avalon-drv/avalon-drv-dev.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-ioctl.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-ioctl.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-mmap.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-mmap.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-sg-buf.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-sg-buf.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-util.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-util.h create mode 100644 drivers/staging/avalon-drv/aval
[PATCH RFC 2/2] staging: avalon-drv: Avalon DMA driver
This is sample implementation of a driver that uses "avalon-dma" driver interface to perform data transfers between on-chip and system memory in devices using Avalon-MM DMA Interface for PCIe design. Companion user-level tool could be found at g...@github.com:a-gordeev/avalon-drv-tool.git CC: Michael Chen CC: de...@driverdev.osuosl.org CC: dmaeng...@vger.kernel.org Signed-off-by: Alexander Gordeev --- drivers/staging/Kconfig | 2 + drivers/staging/Makefile | 1 + drivers/staging/avalon-drv/Kconfig| 34 + drivers/staging/avalon-drv/Makefile | 14 + drivers/staging/avalon-drv/avalon-drv-dev.c | 193 ++ drivers/staging/avalon-drv/avalon-drv-ioctl.c | 137 drivers/staging/avalon-drv/avalon-drv-ioctl.h | 12 + drivers/staging/avalon-drv/avalon-drv-mmap.c | 93 +++ drivers/staging/avalon-drv/avalon-drv-mmap.h | 12 + .../staging/avalon-drv/avalon-drv-sg-buf.c| 132 .../staging/avalon-drv/avalon-drv-sg-buf.h| 26 + drivers/staging/avalon-drv/avalon-drv-util.c | 54 ++ drivers/staging/avalon-drv/avalon-drv-util.h | 12 + drivers/staging/avalon-drv/avalon-drv-xfer.c | 655 ++ drivers/staging/avalon-drv/avalon-drv-xfer.h | 24 + drivers/staging/avalon-drv/avalon-drv.h | 22 + include/uapi/linux/avalon-drv-ioctl.h | 30 + 17 files changed, 1453 insertions(+) create mode 100644 drivers/staging/avalon-drv/Kconfig create mode 100644 drivers/staging/avalon-drv/Makefile create mode 100644 drivers/staging/avalon-drv/avalon-drv-dev.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-ioctl.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-ioctl.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-mmap.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-mmap.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-sg-buf.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-sg-buf.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-util.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-util.h create mode 100644 drivers/staging/avalon-drv/avalon-drv-xfer.c create mode 100644 drivers/staging/avalon-drv/avalon-drv-xfer.h create mode 100644 drivers/staging/avalon-drv/avalon-drv.h create mode 100644 include/uapi/linux/avalon-drv-ioctl.h diff --git a/drivers/staging/Kconfig b/drivers/staging/Kconfig index 31c732ececd1..6a4e9fdae9f6 100644 --- a/drivers/staging/Kconfig +++ b/drivers/staging/Kconfig @@ -122,4 +122,6 @@ source "drivers/staging/isdn/Kconfig" source "drivers/staging/avalon-dma/Kconfig" +source "drivers/staging/avalon-drv/Kconfig" + endif # STAGING diff --git a/drivers/staging/Makefile b/drivers/staging/Makefile index eb974cac85d3..e09e1aea1506 100644 --- a/drivers/staging/Makefile +++ b/drivers/staging/Makefile @@ -51,3 +51,4 @@ obj-$(CONFIG_FIELDBUS_DEV) += fieldbus/ obj-$(CONFIG_KPC2000) += kpc2000/ obj-$(CONFIG_ISDN_CAPI)+= isdn/ obj-$(CONFIG_AVALON_DMA) += avalon-dma/ +obj-$(CONFIG_AVALON_DRV) += avalon-drv/ diff --git a/drivers/staging/avalon-drv/Kconfig b/drivers/staging/avalon-drv/Kconfig new file mode 100644 index ..18c43a65b6f5 --- /dev/null +++ b/drivers/staging/avalon-drv/Kconfig @@ -0,0 +1,34 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA driver +# +# Author: Alexander Gordeev +# +config AVALON_DRV + depends on AVALON_DMA + tristate "Avalon DMA driver" + help + This selects a driver for Avalon DMA + +config AVALON_DRV_PCI_VENDOR_ID + hex "PCI vendor ID" + default "0x1172" + +config AVALON_DRV_PCI_DEVICE_ID + hex "PCI device ID" + default "0xe003" + +config AVALON_DRV_PCI_BAR + int "PCI device BAR the Avalon DMA controller is mapped to" + range 0 5 + default 0 + +config AVALON_DRV_PCI_MSI_COUNT_ORDER + int "Count of MSIs the PCI device provides (order)" + range 0 5 + default 5 + +config AVALON_DRV_PCI_MSI_VECTOR + int "MSI vector the Avalon DMA controller uses in multi-MSI mode" + range 0 31 + default 0 diff --git a/drivers/staging/avalon-drv/Makefile b/drivers/staging/avalon-drv/Makefile new file mode 100644 index ..2e0f5f46488e --- /dev/null +++ b/drivers/staging/avalon-drv/Makefile @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Avalon DMA driver +# +# Author: Alexander Gordeev +# +obj-$(CONFIG_AVALON_DRV) += avalon-drv.o + +avalon-drv-objs := avalon-drv-dev.o \ + avalon-drv-ioctl.o \ + avalon-drv-mmap.o \ + avalon-drv-sg-buf.o \ + avalon-drv-xfer.o \ + avalon-drv-util.o diff --git a/drivers/staging/avalon-drv/avalon-drv-dev.c b/drivers/staging/avalon-drv/aval