| Age | Commit message (Collapse) | Author |
|---|
|
This patch adds dump of registers with debugfs for HIP08
and HIP09 DMA driver.
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220830062251.52993-7-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
The HiSilicon IP08 and HiSilicon IP09 are DMA iEPs, they
have the same pci device id but different pci revision.
Unfortunately, they have different register layouts, so
the origin driver cannot run on HiSilicon IP09 correctly.
This patch enables the driver to adapt to HiSilicon IP09.
HiSilicon IP09 offers 4 channels, each channel has a send
queue, a complete queue and an interrupt to help to do tasks.
This DMA engine can do memory copy between memory blocks.
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220830062251.52993-6-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
readl_relaxed_poll_timeout() uses magic numbers 10 and 1000, which
indicate maximum time to sleep between reads in us and timeout in us,
respectively.
Use macros HISI_DMA_POLL_Q_STS_DELAY_US and
HISI_DMA_POLL_Q_STS_TIME_OUT_US instead of these two numbers.
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220830062251.52993-5-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
When we get a DMA channel and try to use it in multiple threads it
will cause oops and hanging the system.
% echo 100 > /sys/module/dmatest/parameters/threads_per_chan
% echo 100 > /sys/module/dmatest/parameters/iterations
% echo 1 > /sys/module/dmatest/parameters/run
[383493.327077] Unable to handle kernel paging request at virtual
address dead000000000108
[383493.335103] Mem abort info:
[383493.335103] ESR = 0x96000044
[383493.335105] EC = 0x25: DABT (current EL), IL = 32 bits
[383493.335107] SET = 0, FnV = 0
[383493.335108] EA = 0, S1PTW = 0
[383493.335109] FSC = 0x04: level 0 translation fault
[383493.335110] Data abort info:
[383493.335111] ISV = 0, ISS = 0x00000044
[383493.364739] CM = 0, WnR = 1
[383493.367793] [dead000000000108] address between user and kernel
address ranges
[383493.375021] Internal error: Oops: 96000044 [#1] PREEMPT SMP
[383493.437574] CPU: 63 PID: 27895 Comm: dma0chan0-copy2 Kdump:
loaded Tainted: GO 5.17.0-rc4+ #2
[383493.457851] pstate: 204000c9 (nzCv daIF +PAN -UAO -TCO -DIT
-SSBS BTYPE=--)
[383493.465331] pc : vchan_tx_submit+0x64/0xa0
[383493.469957] lr : vchan_tx_submit+0x34/0xa0
This occurs because the transmission timed out, and that's due
to data race. Each thread rewrite channels's descriptor as soon as
device_issue_pending is called. It leads to the situation that
the driver thinks that it uses the right descriptor in interrupt
handler while channels's descriptor has been changed by other
thread. The descriptor which in fact reported interrupt will not
be handled any more, as well as its tx->callback.
That's why timeout reports.
With current fixes channels' descriptor changes it's value only
when it has been used. A new descriptor is acquired from
vc->desc_issued queue that is already filled with descriptors
that are ready to be sent. Threads have no direct access to DMA
channel descriptor. In case of channel's descriptor is busy, try
to submit to HW again when a descriptor is completed. In this case,
vc->desc_issued may be empty when hisi_dma_start_transfer is called,
so delete error reporting on this. Now it is just possible to queue
a descriptor for further processing.
Fixes: e9f08b65250d ("dmaengine: hisilicon: Add Kunpeng DMA engine support")
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220830062251.52993-4-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
After completion of data transfer of one or multiple descriptors,
the completion status and the current head pointer to submission
queue are written into the CQ and interrupt can be generated to
inform the software. In interrupt process CQ is read and cq_head
is updated.
hisi_dma_irq updates cq_head only when the completion status is
success. When an abnormal interrupt reports, cq_head will not update
which will cause subsequent interrupt processes read the error CQ
and never report the correct status.
This patch updates cq_head whenever CQ is accessed.
Fixes: e9f08b65250d ("dmaengine: hisilicon: Add Kunpeng DMA engine support")
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220830062251.52993-3-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
When hisi_dma is unloaded or unbinded, all of channels should be
disabled. This patch disables DMA channels when driver is unloaded
or unbinded.
Fixes: e9f08b65250d ("dmaengine: hisilicon: Add Kunpeng DMA engine support")
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220830062251.52993-2-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
Remove the loaded hisi_dma driver and reload it, the driver fails
to work properly. The following error is reported in the kernel log:
[ 1475.597609] hisi_dma 0000:7b:00.0: Failed to allocate MSI vectors!
[ 1475.604915] hisi_dma: probe of 0000:7b:00.0 failed with error -28
As noted in "The MSI Driver Guide HOWTO"[1], the number of MSI
interrupt must be a power of two. The Kunpeng DMA driver allocates 30
MSI interrupts. As a result, no space left on device is reported
when the driver is reloaded and allocates interrupt vectors from the
interrupt domain.
This patch changes the number of interrupt vectors allocated by
hisi_dma driver to 32 to avoid this problem.
[1] https://www.kernel.org/doc/html/latest/PCI/msi-howto.html
Fixes: e9f08b65250d ("dmaengine: hisilicon: Add Kunpeng DMA engine support")
Signed-off-by: Jie Hai <haijie1@huawei.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20220216072101.34473-1-haijie1@huawei.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
The wrappers in include/linux/pci-dma-compat.h should go away.
pci_set_dma_mask()/pci_set_consistent_dma_mask() should be
replaced with dma_set_mask()/dma_set_coherent_mask(),
and use dma_set_mask_and_coherent() for both.
Signed-off-by: Qing Wang <wangqing@vivo.com>
Link: https://lore.kernel.org/r/1633663733-47199-5-git-send-email-wangqing@vivo.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
When using 'pcim_enable_device()', 'pci_alloc_irq_vectors()' is
auto-magically a managed function.
It is useless (but harmless) to record an action to explicitly call
'pci_free_irq_vectors()'.
So keep things simple, comment why and how these resources are freed, axe
some useless code and save some memory.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/4f8932e2d0d8d092bf60272511100030e013bc72.1623875508.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
Running in hardIRQ, disabling IRQ is redundant since hardIRQ has disabled
IRQ. This patch removes the irqsave and irqstore to save some instruction
cycles.
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Acked-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20201027215252.25820-8-song.bao.hua@hisilicon.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
Make use of the struct_size() helper instead of an open-coded version
in order to avoid any potential type mistakes.
This code was detected with the help of Coccinelle and, audited and
fixed manually.
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
Link: https://lore.kernel.org/r/20200617211135.GA8660@embeddedor
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
|
This patch adds a driver for HiSilicon Kunpeng DMA engine. This DMA engine
which is an PCIe iEP offers 30 channels, each channel has a send queue, a
complete queue and an interrupt to help to do tasks. This DMA engine can do
memory copy between memory blocks or between memory and device buffer.
Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com>
Signed-off-by: Zhenfa Qiu <qiuzhenfa@hisilicon.com>
Link: https://lore.kernel.org/r/1579155057-80523-1-git-send-email-wangzhou1@hisilicon.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
|
