User experienced performance dip on measuring IOPS while EDIF
enabled. During I/O time, driver uses dma_pool_zalloc() call to allocate a
chunk of memory. This call contains a lock behind the scene which
contribute to lock contention. Save the allocated memory for reuse and
avoid the lock.
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
There is no functional change in this patch. VP map resource is renamed
and relocated so it is not viewed as just a target mode resource.
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This code accidentally returns success instead of -ENOMEM.
Fixes: 7cc7646b4b ("scsi: libsas: Factor out sas_ata_add_dev()")
Link: https://lore.kernel.org/r/Y7asLxzVwQ56G+ya@kili
Signed-off-by: Dan Carpenter <error27@gmail.com>
Reviewed-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Currently, it used old rss size to get current tc mode. As a result, the
rss size is updated, but the tc mode is still configured based on the old
rss size.
So this patch fixes it by using the new rss size in both process.
Fixes: 93969dc14f ("net: hns3: refactor VF rss init APIs with new common rss init APIs")
Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Hao Lan <lanhao@huawei.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/20230110115359.10163-1-lanhao@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are still single reports of systems where ASPM incompatibilities
cause tx timeouts. It's not clear whom to blame, so let's disable
ASPM in case of a tx timeout.
v2:
- add one-time warning for informing the user
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://lore.kernel.org/r/92369a92-dc32-4529-0509-11459ba0e391@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Bean Huo <beanhuo@iokpp.de> says:
These patches are to fix several compilation warnings introduced by my
commit: 6ff265fc5e ("scsi: ufs: core: bsg: Add advanced RPMB support
in ufs_bsg"), please consider this patch series for the next your
merge window. Apologies for this!!
Link: https://lore.kernel.org/r/20230108224057.354438-1-beanhuo@iokpp.de
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fix the following sparse endianness warning:
"sparse warnings: drivers/ufs/core/ufs_bsg.c:91:25: sparse: sparse: cast to
restricted __be16."
For consistency with endianness annotations of other UFS data structures,
change __u16/32 to __be16/32 in UFS ARPMB data structures.
Fixes: 6ff265fc5e ("scsi: ufs: core: bsg: Add advanced RPMB support in ufs_bsg")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Bean Huo <beanhuo@micron.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In 'include/ufs/ufshcd.h' file, 'enum dma_data_direction' will be used,
which is defined in linux/dma-direction.h, however, this header file is not
included in ufshcd.h, thus causing the following compilation warning:
"warning: ‘enum dma_data_direction’ declared inside parameter list will not
be visible outside of this definition or declaration"
Fix this warning by including 'linux/dma-direction.h'.
Fixes: 6ff265fc5e ("scsi: ufs: core: bsg: Add advanced RPMB support in ufs_bsg")
Reported-by: Xiaosen He <quic_xiaosenh@quicinc.com>
Reported-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Bean Huo <beanhuo@micron.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Compilation complains that two possible variables are used without
initialization:
drivers/ufs/core/ufs_bsg.c:112:6: warning: variable 'sg_cnt' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
drivers/ufs/core/ufs_bsg.c:112:6: warning: variable 'sg_list' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
Fix both warnings by adding initialization with sg_cnt = 0, sg_list = NULL.
Fixes: 6ff265fc5e ("scsi: ufs: core: bsg: Add advanced RPMB support in ufs_bsg")
Signed-off-by: Bean Huo <beanhuo@micron.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Xiaosen He <quic_xiaosenh@quicinc.com>
Reviewed-by: Alim Akhtar <alim.akhtar@samsung.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
According to Documentation/filesystems/sysfs.rst, the show() callback
function of kobject attributes should use sysfs_emit() instead of the
sprintf() family of functions.
Issue identified using the coccinelle device_attr_show.cocci script.
Link: https://lore.kernel.org/r/Y5JE/xI2NNbnox/A@qemulion
Signed-off-by: Deepak R Varma <drv@mailo.com>
Acked-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Nilesh Javali <njavali@marvell.com> says:
Please apply the miscellaneous qla2xxx driver bug fixes to the SCSI
tree at your earliest convenience.
Link: https://lore.kernel.org/r/20221219110748.7039-1-njavali@marvell.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> says:
This series adds HS-G4 support to the Qcom UFS driver and PHY driver.
The newer Qcom platforms support configuring the UFS controller and
PHY in dual gears (i.e., controller/PHY can be configured to run in
two gear speeds). This is accomplished by adding two different PHY
init sequences to the PHY driver and the UFS driver requesting the one
that's required based on the platform configuration.
Initially the ufs-qcom driver will use the default gear G2 for
enumerating the UFS device. Afer enumeration, the max gear supported
by both the controller and device would be found out and that will be
used thereafter. But for using the max gear after enumeration, the
ufs-qcom driver requires the UFS device to be reinitialized. For this
purpose, a separate quirk has been introduced in the UFS core along
with a callback and those will be used by the ufs-qcom driver.
This series has been tested on following platforms:
* Qcom RB5 development platform powered by SM8250 SoC
* SM8450 based dev board
* Qdrive3/sa8540p-ride board based on SC8280XP (derivative)
Link: https://lore.kernel.org/r/20221222141001.54849-1-manivannan.sadhasivam@linaro.org
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Qcom UFS drivers are left unmaintained till now. I'd like to step up to
maintain the drivers and the binding.
Acked-by: Bjorn Andersson <andersson@kernel.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Starting from Qcom UFS version 4.0, vendor specific REG_UFS_PARAM0 register
can be used to determine the maximum gear supported by the controller.
Suggested-by: Can Guo <quic_cang@quicinc.com>
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Starting from Qualcomm UFS version 4, the UFS device needs to be
reinitialized after switching to maximum gear by the UFS core. Hence, add
support for it by enabling the UFSHCD_QUIRK_REINIT_AFTER_MAX_GEAR_SWITCH
quirk, implementing reinit_notify() callback and using the agreed gear
speed for setting the PHY mode.
Suggested-by: Can Guo <quic_cang@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In the preparation of adding support for new gears, move the logic that
finds the gear for each platform to a new function. This helps with code
readability and also allows the logic to be used in other places of the
driver in future.
While at it, make it clear that this driver only supports symmetric gear
setting (hs_tx_gear == hs_rx_gear).
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some platforms like Qcom, requires the UFS device to be reinitialized after
switching to maximum gear speed. So add support for that in UFS core by
introducing a new quirk (UFSHCD_CAP_REINIT_AFTER_MAX_GEAR_SWITCH) and doing
the reinitialization, if the quirk is enabled by the controller driver.
Suggested-by: Can Guo <quic_cang@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
reinit_notify() callback can be used by the UFS controller drivers to
perform changes required for UFSHCD reinit that can happen during max gear
switch.
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
On newer UFS revisions, the register at offset 0xD0 is called,
REG_UFS_PARAM0. Since the existing register, RETRY_TIMER_REG is not used
anywhere, it is safe to use the new name.
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Reviewed-by: Asutosh Das <quic_asutoshd@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Make use of dev_err_probe() for printing the probe error.
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Reviewed-by: Asutosh Das <quic_asutoshd@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Use bitfield macros where appropriate to simplify the driver.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Asutosh Das <quic_asutoshd@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In the reset assert and deassert callbacks, the supplied "id" is not used
at all and only the HBA reset is performed all the time. So there is no
reason to use a WARN_ON() on the "id".
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Reviewed-by: Asutosh Das <quic_asutoshd@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
A goto statement in an error path is useful if the function needs to do
cleanup other than returning the error code. But in this driver, goto
statements are used for just returning the error code in many places. This
really makes it hard to read the code.
Get rid of those goto statements and just return the error code directly.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Asutosh Das <quic_asutoshd@quicinc.com>
Tested-by: Andrew Halaney <ahalaney@redhat.com> # Qdrive3/sa8540p-ride
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Currently the nfsd-client shrinker is registered and unregistered at
the time the nfsd module is loaded and unloaded. The problem with this
is the shrinker is being registered before all of the relevant fields
in nfsd_net are initialized when nfsd is started. This can lead to an
oops when memory is low and the shrinker is called while nfsd is not
running.
This patch moves the register/unregister of nfsd-client shrinker from
module load/unload time to nfsd startup/shutdown time.
Fixes: 44df6f439a ("NFSD: add delegation reaper to react to low memory condition")
Reported-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
If signal_pending() returns true, schedule_timeout() will not be executed,
causing the waiting task to remain in the wait queue.
Fixed by adding a call to finish_wait(), which ensures that the waiting
task will always be removed from the wait queue.
Fixes: f4e44b3933 ("NFSD: delay unmount source's export after inter-server copy completed.")
Signed-off-by: Xingyuan Mo <hdthky0@gmail.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Make qla_get_iocbs_resource() static to fix the warning:
>> drivers/scsi/qla2xxx/qla_iocb.c:3820:5: warning: no previous prototype for
>> 'qla_get_iocbs_resource' [-Wmissing-prototypes]
3820 | int qla_get_iocbs_resource(struct srb *sp)
| ^~~~~~~~~~~~~~~~~~~~~~
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Residual underrun is not an interface error, hence no need to increment
that count.
Fixes: dbf1f53cfd ("scsi: qla2xxx: Implementation to get and manage host, target stats and initiator port")
Cc: stable@vger.kernel.org
Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If after an adapter reset the appearance of link is not recovered, the
devices are not rediscovered. This is result of a race condition between
adapter reset (abort_isp) and the topology scan. During adapter reset, the
ABORT_ISP_ACTIVE flag is set. Topology scan usually occurred after adapter
reset. In this case, the topology scan came earlier than usual where it
ran into problem due to ABORT_ISP_ACTIVE flag was still set.
kernel: qla2xxx [0000:13:00.0]-1005:1: Cmd 0x6a aborted with timeout since ISP Abort is pending
kernel: qla2xxx [0000:13:00.0]-28a0:1: MBX_GET_PORT_NAME failed, No FL Port.
kernel: qla2xxx [0000:13:00.0]-286b:1: qla2x00_configure_loop: exiting normally. local port wwpn 51402ec0123d9a80 id 012300)
kernel: qla2xxx [0000:13:00.0]-8017:1: ADAPTER RESET SUCCEEDED nexus=1:0:15.
Allow adapter reset to complete before any scan can start.
Cc: stable@vger.kernel.org
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
FCF_ASYNC_SENT flag is used in session management. This flag is cleared in
task management path by accident. Remove unintended flag clearing.
Fixes: 388a49959e ("scsi: qla2xxx: Fix panic from use after free in qla2x00_async_tm_cmd")
Cc: stable@vger.kernel.org
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If a login failed due to low FW resources, the session can stall and will
not be connected. Reset session state to allow relogin logic to redrive
the connection.
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In large environment, it is possible to experience command timeout and
escalation of path recovery. Currently the driver does not track the number
of exchanges/commands sent to FW. If there is a delay for commands at the
head of the queue, then this will create back pressure for commands at the
back of the queue.
Check for exchange availability before command submission.
Fixes: 89c72f4245 ("scsi: qla2xxx: Add IOCB resource tracking")
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The following message and call trace was seen with debug kernels:
DMA-API: qla2xxx 0000:41:00.0: device driver failed to check map
error [device address=0x00000002a3ff38d8] [size=1024 bytes] [mapped as
single]
WARNING: CPU: 0 PID: 2930 at kernel/dma/debug.c:1017
check_unmap+0xf42/0x1990
Call Trace:
debug_dma_unmap_page+0xc9/0x100
qla_nvme_ls_unmap+0x141/0x210 [qla2xxx]
Remove DMA mapping from the driver altogether, as it is already done by FC
layer. This prevents the warning.
Fixes: c85ab7d9e2 ("scsi: qla2xxx: Fix missed DMA unmap for NVMe ls requests")
Cc: stable@vger.kernel.org
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
User experienced symptoms of adapter failure in NPIV environment. NPIV
hosts were allowed to trigger chip reset back to back due to NPIV link
state being slow to come online.
Fix link failure in NPIV environment by removing NPIV host from directly
being able to perform chip reset.
kernel: qla2xxx [0000:04:00.1]-6009:261: Loop down - aborting ISP.
kernel: qla2xxx [0000:04:00.1]-6009:262: Loop down - aborting ISP.
kernel: qla2xxx [0000:04:00.1]-6009:281: Loop down - aborting ISP.
kernel: qla2xxx [0000:04:00.1]-6009:285: Loop down - aborting ISP
Fixes: 0d6e61bc6a ("[SCSI] qla2xxx: Correct various NPIV issues.")
Cc: stable@vger.kernel.org
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
CT Ping and ELS cmds fail for NVMe targets. Check if port is online before
sending ELS instead of sending login.
Cc: stable@vger.kernel.org
Signed-off-by: Shreyas Deodhar <sdeodhar@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add driver code for Unisoc UFS host controller, along with UFS
initialization.
Signed-off-by: Zhe Wang <zhe.wang1@unisoc.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Magnus Karlsson says:
====================
This is a patch set of various performance improvements, fixes, and
the introduction of more than one XDP program to the xsk selftests
framework so we can test more things in the future such as upcoming
multi-buffer and metadata support for AF_XDP. The new programs just
reuse the framework that all the other eBPF selftests use. The new
feature is used to implement one new test that does XDP_DROP on every
other packet. More tests using this will be added in future commits.
Contents:
* The run-time of the test suite is cut by 10x when executing the
tests on a real NIC, by only attaching the XDP program once per mode
tested, instead of once per test program.
* Over 700 lines of code have been removed. The xsk.c control file was
moved straight over from libbpf when the xsk support was deprecated
there. As it is now not used as library code that has to work with
all kinds of versions of Linux, a lot of code could be dropped or
simplified.
* Add a new command line option "-d" that can be used when a test
fails and you want to debug it with gdb or some other debugger. The
option creates the two veth netdevs and prints them to the screen
without deleting them afterwards. This way these veth netdevs can be
used when running xskxceiver in a debugger.
* Implemented the possibility to load external XDP programs so we can
have more than the default one. This feature is used to implement a
test where every other packet is dropped. Good exercise for the
recycling mechanism of the xsk buffer pool used in zero-copy mode.
* Various clean-ups and small fixes in patches 1 to 5. None of these
fixes has any impact on the correct execution of the tests when they
pass, though they can be irritating when a test fails. IMHO, they do
not need to go to bpf as they will not fix anything there. The first
version of patches 1, 2, and 4 where previously sent to bpf, but has
now been included here.
v2 -> v3:
* Fixed compilation error for llvm [David]
* Made the function xsk_is_in_drv_mode(ifobj) more generic by changing
it to xsk_is_in_mode(ifobj, xdp_mode) [Maciej]
* Added Maciej's acks to all the patches
v1 -> v2:
* Fixed spelling error in commit message of patch #6 [Björn]
* Added explanation on why it is safe to use C11 atomics in patch #7
[Daniel]
* Put all XDP programs in the same file so that adding more XDP
programs to xskxceiver.c becomes more scalable in patches #11 and
#12 [Maciej]
* Removed more dead code in patch #8 [Maciej]
* Removed stale %s specifier in error print, patch #9 [Maciej]
* Changed name of XDP_CONSUMES_SOME_PACKETS to XDP_DROP_HALF to
hopefully make it clearer [Maciej]
* ifobj_rx and ifobj_tx name changes in patch #13 [Maciej]
* Simplified XDP attachment code in patch #15 [Maciej]
Patches:
1-5: Small fixes and clean-ups
6: New convenient debug option when using a debugger such as gdb
7-8: Removal of unnecessary code
9: Add the ability to load external XDP programs
10-11: Removal of more unnecessary code
12: Implement a new test where every other packet is XDP_DROP:ed
13: Unify the thread dispatching code
14-15: Simplify the way tests are written when using custom packet_streams
or custom XDP programs
Thanks: Magnus
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Implement automatic switching of XDP programs and execution modes if
needed by a test. This makes it much simpler to write a test as it
only has to say what XDP program it needs if it is not the default
one. This also makes it possible to remove the initial explicit
attachment of the XDP program as well as the explicit mode switch in
the code. These are now handled by the same code that just checks if a
switch is necessary, so no special cases are needed.
The default XDP program for all tests is one that sends all packets to
the AF_XDP socket. If you need another one, please use the new
function test_spec_set_xdp_prog() to specify what XDP programs and
maps to use for this test.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-16-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Automatically restore the default packet stream if needed at the end
of each test. This so that test writers do not forget to do this.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-15-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Make the thread dispatching code common by unifying the dual and
single thread dispatcher code. This so we do not have to add code in
two places in upcoming commits.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-14-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add a new test where some of the packets are not passed to the AF_XDP
socket and instead get a XDP_DROP verdict. This is important as it
tests the recycling mechanism of the buffer pool. If a packet is not
sent to the AF_XDP socket, the buffer the packet resides in is instead
recycled so it can be used again without the round-trip to user
space. The test introduces a new XDP program that drops every other
packet.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-13-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Get rid of the built-in XDP program that was part of the old libbpf
code in xsk.c and replace it with an eBPF program build using the
framework by all the other bpf selftests. This will form the base for
adding more programs in later commits.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-12-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Remove unnecessary code in the control path. This is located in the
file xsk.c that was moved from libbpf when the xsk support there was
deprecated. Some of the code there is not needed anymore as the
selftests are only guaranteed to run on the kernel it is shipped
with. Therefore, all the code that has to deal with compatibility of
older kernels can be dropped and also any other function that is not
of any use for the tests.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-11-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Load and attach the XDP program only once per XDP mode that is being
executed. Today, the XDP program is loaded and attached for every
test, then unloaded, which takes a long time on real NICs, since they
have to reconfigure their HW, in contrast to veth. The test suite now
completes in 21 seconds, instead of 207 seconds previously on my
machine. This is a speed-up of around 10x.
This is accomplished by moving the XDP loading from the worker threads
to the main thread and replacing the XDP loading interfaces of xsk.c
that was taken from the xsk support in libbpf, with something more
explicit that is more useful for these tests. Instead, the relevant
file descriptors and ifindexes are just passed down to the new
functions.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20230111093526.11682-10-magnus.karlsson@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
