syzbot is reporting hung task at wdm_flush() [1], for there is a circular
dependency that wdm_flush() from flip_close() for /dev/cdc-wdm0 forever
waits for /dev/raw-gadget to be closed while close() for /dev/raw-gadget
cannot be called unless close() for /dev/cdc-wdm0 completes.
Tetsuo Handa considered that such circular dependency is an usage error [2]
which corresponds to an unresponding broken hardware [3]. But Alan Stern
responded that we should be prepared for such hardware [4]. Therefore,
this patch changes wdm_flush() to use wait_event_interruptible_timeout()
which gives up after 30 seconds, for hardware that remains silent must be
ignored. The 30 seconds are coming out of thin air.
Changing wait_event() to wait_event_interruptible_timeout() makes error
reporting from close() syscall less reliable. To compensate it, this patch
also implements wdm_fsync() which does not use timeout. Those who want to
be very sure that data has gone out to the device are now advised to call
fsync(), with a caveat that fsync() can return -EINVAL when running on
older kernels which do not implement wdm_fsync().
This patch also fixes three more problems (listed below) found during
exhaustive discussion and testing.
Since multiple threads can concurrently call wdm_write()/wdm_flush(),
we need to use wake_up_all() whenever clearing WDM_IN_USE in order to
make sure that all waiters are woken up. Also, error reporting needs
to use fetch-and-clear approach in order not to report same error for
multiple times.
Since wdm_flush() checks WDM_DISCONNECTING, wdm_write() should as well
check WDM_DISCONNECTING.
In wdm_flush(), since locks are not held, it is not safe to dereference
desc->intf after checking that WDM_DISCONNECTING is not set [5]. Thus,
remove dev_err() from wdm_flush().
[1] https://syzkaller.appspot.com/bug?id=e7b761593b23eb50855b9ea31e3be5472b711186
[2] https://lkml.kernel.org/r/27b7545e-8f41-10b8-7c02-e35a08eb1611@i-love.sakura.ne.jp
[3] https://lkml.kernel.org/r/79ba410f-e0ef-2465-b94f-6b9a4a82adf5@i-love.sakura.ne.jp
[4] https://lkml.kernel.org/r/20200530011040.GB12419@rowland.harvard.edu
[5] https://lkml.kernel.org/r/c85331fc-874c-6e46-a77f-0ef1dc075308@i-love.sakura.ne.jp
Reported-by: syzbot <syzbot+854768b99f19e89d7f81@syzkaller.appspotmail.com>
Cc: stable <stable@vger.kernel.org>
Co-developed-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Oliver Neukum <oneukum@suse.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Link: https://lore.kernel.org/r/20200928141755.3476-1-penguin-kernel@I-love.SAKURA.ne.jp
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The old usb_control_msg() let the caller handle the error and also did not
account for partial reads. Since these are now considered harmful, move the
driver over to usb_control_msg_recv/send() calls.
Added small note about why set_registers() can't be used to substitute
set_register().
Signed-off-by: Petko Manolov <petko.manolov@konsulko.com>
Acked-by: David S. Miller <davem@davemloft.net>
Link: https://lore.kernel.org/r/20200927124909.16380-2-petko.manolov@konsulko.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The old usb_control_msg() let the caller handle the error and also did not
account for partial reads. Since these are now considered harmful, move the
driver over to usb_control_msg_recv/send() calls.
Signed-off-by: Petko Manolov <petko.manolov@konsulko.com>
Acked-by: David S. Miller <davem@davemloft.net>
Link: https://lore.kernel.org/r/20200927124909.16380-3-petko.manolov@konsulko.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It's not an error if the mode can't be entered because
another mode is already active, so no longer printing an
error message if that happens.
Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://lore.kernel.org/r/20200928133324.48841-1-heikki.krogerus@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Description based on one by Yasushi Asano:
According to 6.7.22 A-UUT “Device No Response” for connection timeout
of USB OTG and EH automated compliance plan v1.2, enumeration failure
has to be detected within 30 seconds. However, the old and new
enumeration schemes each make a total of 12 attempts, and each attempt
can take 5 seconds to time out, so the PET test fails.
This patch adds a new Kconfig option (CONFIG_USB_FEW_INIT_RETRIES);
when the option is set all the initialization retry loops except the
outermost are reduced to a single iteration. This reduces the total
number of attempts to four, allowing Linux hosts to pass the PET test.
The new option is disabled by default to preserve the existing
behavior. The reduced number of retries may fail to initialize a few
devices that currently do work, but for the most part there should be
no change. And in cases where the initialization does fail, it will
fail much more quickly.
Reported-and-tested-by: yasushi asano <yazzep@gmail.com>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Link: https://lore.kernel.org/r/20200928152217.GB134701@rowland.harvard.edu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The SET_CONFIG_TRIES macro in hub.c is badly named; it controls the
number of port-initialization retry attempts rather than the number of
Set-Configuration attempts. Furthermore, the USE_NEW_SCHEME macro and
use_new_scheme() function are written in a very confusing manner,
making it almost impossible to figure out exactly what they do or
check that they are correct.
This patch renames SET_CONFIG_TRIES to PORT_INIT_TRIES, removes
USE_NEW_SCHEME entirely, and rewrites use_new_scheme() to be much more
transparent, with added comments explaining how it works. The patch
also pulls the single call site of use_new_scheme() out from the
Get-Descriptor retry loop (where it returns the same value each time)
and renames the local variable used to store the result.
The overall effect is a minor cleanup. However, there is one
functional change: If the "use_both_schemes" module parameter isn't
set (by default it is set), the existing code does only two retry
iterations. After this patch it will always perform four, regardless
of the parameter's value.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Link: https://lore.kernel.org/r/20200928152050.GA134701@rowland.harvard.edu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a change in the MMU notifier sequence number forces user_mem_abort()
to return early when attempting to handle a stage-2 fault, we return
uninitialised stack to kvm_handle_guest_abort(), which could potentially
result in the injection of an external abort into the guest or a spurious
return to userspace. Neither or these are what we want to do.
Initialise 'ret' to 0 in user_mem_abort() so that bailing due to a
change in the MMU notrifier sequence number is treated as though the
fault was handled.
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Alexandru Elisei <alexandru.elisei@arm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Cc: Gavin Shan <gshan@redhat.com>
Cc: Alexandru Elisei <alexandru.elisei@arm.com>
Link: https://lore.kernel.org/r/20200930102442.16142-1-will@kernel.org
Alex pointed out that we don't pass a level hint to the TLBI instruction
when handling a stage-2 permission fault, even though the walker does
at some point have the level information in its hands.
Rework stage2_update_leaf_attrs() so that it can optionally return the
level of the updated pte to its caller, which can in turn be used to
provide the correct TLBI level hint.
Reported-by: Alexandru Elisei <alexandru.elisei@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Alexandru Elisei <alexandru.elisei@arm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Cc: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/595cc73e-636e-8b3a-f93a-b4e9fb218db8@arm.com
Link: https://lore.kernel.org/r/20200930131801.16889-1-will@kernel.org
As warned with make htmldocs:
.../Documentation/virt/kvm/devices/vcpu.rst:70: WARNING: Malformed table.
Text in column margin in table line 2.
======= ======================================================
-ENODEV: PMUv3 not supported or GIC not initialized
-ENXIO: PMUv3 not properly configured or in-kernel irqchip not
configured as required prior to calling this attribute
-EBUSY: PMUv3 already initialized
-EINVAL: Invalid filter range
======= ======================================================
The ':' character for two lines are above the size of the column.
Besides that, other tables at the file doesn't use ':', so
just drop them.
While here, also fix this warning also introduced at the same patch:
.../Documentation/virt/kvm/devices/vcpu.rst:88: WARNING: Block quote ends without a blank line; unexpected unindent.
By marking the C code as a literal block.
Fixes: 8be86a5eec ("KVM: arm64: Document PMU filtering API")
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Link: https://lore.kernel.org/r/b5385dd0213f1f070667925bf7a807bf5270ba78.1601616399.git.mchehab+huawei@kernel.org
Nathan reports that building the new mokvar table code for 32-bit
ARM fails with errors such as
error: implicit declaration of function 'early_memunmap'
error: implicit declaration of function 'early_memremap'
This is caused by the lack of an explicit #include of the appropriate
header, and ARM apparently does not inherit that inclusion via another
header file. So add the #include.
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Replace commas with semicolons. What is done is essentially described by
the following Coccinelle semantic patch (http://coccinelle.lip6.fr/):
// <smpl>
@@ expression e1,e2; @@
e1
-,
+;
e2
... when any
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace commas with semicolons. What is done is essentially described by
the following Coccinelle semantic patch (http://coccinelle.lip6.fr/):
// <smpl>
@@ expression e1,e2; @@
e1
-,
+;
e2
... when any
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace commas with semicolons. What is done is essentially described by
the following Coccinelle semantic patch (http://coccinelle.lip6.fr/):
// <smpl>
@@ expression e1,e2; @@
e1
-,
+;
e2
... when any
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace commas with semicolons. What is done is essentially described by
the following Coccinelle semantic patch (http://coccinelle.lip6.fr/):
// <smpl>
@@ expression e1,e2; @@
e1
-,
+;
e2
... when any
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace commas with semicolons. What is done is essentially described by
the following Coccinelle semantic patch (http://coccinelle.lip6.fr/):
// <smpl>
@@ expression e1,e2; @@
e1
-,
+;
e2
... when any
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Currently, we use the jiffies counter as a time source, by staring at
it until a HZ period elapses, and then staring at it again and perform
as many XOR operations as we can at the same time until another HZ
period elapses, so that we can calculate the throughput. This takes
longer than necessary, and depends on HZ, which is undesirable, since
HZ is system dependent.
Let's use the ktime interface instead, and use it to time a fixed
number of XOR operations, which can be done much faster, and makes
the time spent depend on the performance level of the system itself,
which is much more reasonable. To ensure that we have the resolution
we need even on systems with 32 kHz time sources, while not spending too
much time in the benchmark on a slow CPU, let's switch to 3 attempts of
800 repetitions each: that way, we will only misidentify algorithms that
perform within 10% of each other as the fastest if they are faster than
10 GB/s to begin with, which is not expected to occur on systems with
such coarse clocks.
On ThunderX2, I get the following results:
Before:
[72625.956765] xor: measuring software checksum speed
[72625.993104] 8regs : 10169.000 MB/sec
[72626.033099] 32regs : 12050.000 MB/sec
[72626.073095] arm64_neon: 11100.000 MB/sec
[72626.073097] xor: using function: 32regs (12050.000 MB/sec)
After:
[72599.650216] xor: measuring software checksum speed
[72599.651188] 8regs : 10491 MB/sec
[72599.652006] 32regs : 12345 MB/sec
[72599.652871] arm64_neon : 11402 MB/sec
[72599.652873] xor: using function: 32regs (12345 MB/sec)
Link: https://lore.kernel.org/linux-crypto/20200923182230.22715-3-ardb@kernel.org/
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Currently, the XOR module performs its boot time benchmark at core
initcall time when it is built-in, to ensure that the RAID code can
make use of it when it is built-in as well.
Let's defer this to a later stage during the boot, to avoid impacting
the overall boot time of the system. Instead, just pick an arbitrary
implementation from the list, and use that as the preliminary default.
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The 'qm->curr_qm_qp_num' is not initialized, which will result in failure
to write the current_q file.
Signed-off-by: Sihang Chen <chensihang1@hisilicon.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As before, when the ZIP device is too busy to creat a request, it will
return '-EBUSY'. But the crypto process think the '-EBUSY' means a
successful request and wait for its completion.
So replace '-EBUSY' with '-EAGAIN' to show crypto this request is failed.
Fixes: 62c455ca853e("crypto: hisilicon - add HiSilicon ZIP...")
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The zero length input will cause a call trace when use GZIP
decompress like this:
Unable to handle kernel paging request at virtual address
...
lr : get_gzip_head_size+0x7c/0xd0 [hisi_zip]
Judge the input length and return '-EINVAL' when input is invalid.
Fixes: 62c455ca853e("crypto: hisilicon - add HiSilicon ZIP...")
Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
ZIP debug registers aren't cleared even if its driver is removed,
so add a clearing operation when remove driver.
Signed-off-by: Hao Fang <fanghao11@huawei.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch removes a number of unused variables and marks others
as unused in order to silence compiler warnings about them.
Fixes: a8ea8bdd9d ("lib/mpi: Extend the MPI library")
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Tested-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch removes a few ineffectual assignments from the function
crypto_poly1305_setdctxkey.
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Modify the read size to the correct HW random
registers size, 8bit.
The incorrect read size caused and faulty
HW random value.
Signed-off-by: Tomer Maimon <tmaimon77@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Make sure that we call the dma_unmap_sg on the correct scatterlist on
completion with the correct sg_nents.
Use sg_table to managed the DMA mapping and at the same time add the needed
dma_sync calls for the sg_table.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Newer CAAM versions (Era 9+) support 16B IVs. Since for these devices
the HW limitation is no longer present newer version should process the
requests containing 16B IVs directly in hardware without using a fallback.
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Newer CAAM versions (Era 9+) support 16B IVs. Since for these devices
the HW limitation is no longer present newer version should process the
requests containing 16B IVs directly in hardware without using a fallback.
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Newer CAAM versions (Era 9+) support 16B IVs. Since for these devices
the HW limitation is no longer present newer version should process the
requests containing 16B IVs directly in hardware without using a fallback.
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
XTS should not return succes when dealing with block length equal to zero.
This is different than the rest of the skcipher algorithms.
Fixes: 31bb2f0da1 ("crypto: caam - check zero-length input")
Cc: <stable@vger.kernel.org> # v5.4+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
CAAM accelerator only supports XTS-AES-128 and XTS-AES-256 since
it adheres strictly to the standard. All the other key lengths
are accepted and processed through a fallback as long as they pass
the xts_verify_key() checks.
Fixes: 226853ac3e ("crypto: caam/qi2 - add skcipher algorithms")
Cc: <stable@vger.kernel.org> # v4.20+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
CAAM accelerator only supports XTS-AES-128 and XTS-AES-256 since
it adheres strictly to the standard. All the other key lengths
are accepted and processed through a fallback as long as they pass
the xts_verify_key() checks.
Fixes: b189817cf7 ("crypto: caam/qi - add ablkcipher and authenc algorithms")
Cc: <stable@vger.kernel.org> # v4.12+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
CAAM accelerator only supports XTS-AES-128 and XTS-AES-256 since
it adheres strictly to the standard. All the other key lengths
are accepted and processed through a fallback as long as they pass
the xts_verify_key() checks.
Fixes: c6415a6016 ("crypto: caam - add support for acipher xts(aes)")
Cc: <stable@vger.kernel.org> # v4.4+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
A hardware limitation exists for CAAM until Era 9 which restricts
the accelerator to IVs with only 8 bytes. When CAAM has a lower era
a fallback is necessary to process 16 bytes IV.
Fixes: 226853ac3e ("crypto: caam/qi2 - add skcipher algorithms")
Cc: <stable@vger.kernel.org> # v4.20+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
A hardware limitation exists for CAAM until Era 9 which restricts
the accelerator to IVs with only 8 bytes. When CAAM has a lower era
a fallback is necessary to process 16 bytes IV.
Fixes: b189817cf7 ("crypto: caam/qi - add ablkcipher and authenc algorithms")
Cc: <stable@vger.kernel.org> # v4.12+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
A hardware limitation exists for CAAM until Era 9 which restricts
the accelerator to IVs with only 8 bytes. When CAAM has a lower era
a fallback is necessary to process 16 bytes IV.
Fixes: c6415a6016 ("crypto: caam - add support for acipher xts(aes)")
Cc: <stable@vger.kernel.org> # v4.4+
Signed-off-by: Andrei Botila <andrei.botila@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Fix resource leak in error handling.
Signed-off-by: Pavel Machek (CIP) <pavel@denx.de>
Acked-by: John Allen <john.allen@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
With commit 91e81150d3 ("mtd: parsers: bcm63xx: simplify CFE
detection"), we generate a reference to fw_arg3 which is the fourth
firmware/command line argument on MIPS platforms. That symbol is not
exported and would cause a linking failure.
The parser is typically necessary to boot a BCM63xx-based system anyway
so having it be part of the kernel image makes sense, therefore make it
'bool' instead of 'tristate'.
Fixes: 91e81150d3 ("mtd: parsers: bcm63xx: simplify CFE detection")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200929172726.30469-1-f.fainelli@gmail.com
Baikal-T1 Boot Controller provides an access to a RO storages, which are
physically mapped into the SoC MMIO space. In particularly there are
Internal ROM embedded into the SoC with a pre-installed firmware,
externally attached SPI flash (also accessed in the read-only mode) and a
memory region, which mirrors one of them in accordance with the currently
enabled system boot mode (also called Boot ROM).
This commit adds the Internal ROM support to the physmap driver of the MTD
kernel subsystem. The driver will create the Internal ROM MTD as long as
it is defined in the system dts file. The physically mapped SPI flash
region will be used to implement the SPI-mem interface. The mirroring
memory region won't be accessible directly since it's redundant due to
both bootable regions being exposed anyway.
Note we had to create a dedicated code for the ROMs since read from the
corresponding memory regions must be done via the dword-aligned addresses.
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Pavel Parkhomenko <Pavel.Parkhomenko@baikalelectronics.ru>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: linux-mips@vger.kernel.org
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200920111445.21816-1-Sergey.Semin@baikalelectronics.ru
If the transfer had already started and there's no TRB to update, then
there's no need to go through __dwc3_gadget_kick_transfer(). There is
no problem reissuing UPDATE_TRANSFER command. This change just saves
the driver from doing a few operations. This happens when we run out of
TRB and function driver still queues for more requests.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
If we couldn't finish preparing all the TRBs of a request, don't prepare
the next request. Otherwise, the TRBs order will be mixed up and the
controller will process the wrong TRB. This is a corner case where
there's not enough TRBs for a request that needs the extra TRB but
there's still 1 available TRB in the pool.
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
To keep the setting of interrupt-on-completion (IOC) when out of TRBs
consistent and easier to read, the caller of dwc3_prepare_one_trb()
will determine if the TRB must have IOC bit set. This also reduces the
number of times we need to call dwc3_calc_trbs_left(). Note that we only
care about setting IOC from insufficient number of TRBs for SG and not
linear requests (because we don't need to split linear requests).
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
Previously if we run out of TRBs for the last SG entry that requires
an extra TRB, we set interrupt-on-completion (IOC) bit to an already
prepared TRB (i.e. HWO=1). This logic is not clean, and it's not a
typical way to prepare TRB. Also, this prevents showing IOC setup in
tracepoint when __dwc3_prepare_one_trb() is executed. Instead, let's
look ahead when preparing TRB to know whether to set the IOC bit before
the last SG entry. This requires adding a new parameter "must_interrupt"
to dwc3_prepare_one_trb().
Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@kernel.org>
