The value of the register is changed after the task running. A debugfs
file node is added to help users to check the change of register values.
Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Pull VFIO updates from Alex Williamson:
- Introduce new device migration uAPI and implement device specific
mlx5 vfio-pci variant driver supporting new protocol (Jason
Gunthorpe, Yishai Hadas, Leon Romanovsky)
- New HiSilicon acc vfio-pci variant driver, also supporting migration
interface (Shameer Kolothum, Longfang Liu)
- D3hot fixes for vfio-pci-core (Abhishek Sahu)
- Document new vfio-pci variant driver acceptance criteria
(Alex Williamson)
- Fix UML build unresolved ioport_{un}map() functions
(Alex Williamson)
- Fix MAINTAINERS due to header movement (Lukas Bulwahn)
* tag 'vfio-v5.18-rc1' of https://github.com/awilliam/linux-vfio: (31 commits)
vfio-pci: Provide reviewers and acceptance criteria for variant drivers
MAINTAINERS: adjust entry for header movement in hisilicon qm driver
hisi_acc_vfio_pci: Use its own PCI reset_done error handler
hisi_acc_vfio_pci: Add support for VFIO live migration
crypto: hisilicon/qm: Set the VF QM state register
hisi_acc_vfio_pci: Add helper to retrieve the struct pci_driver
hisi_acc_vfio_pci: Restrict access to VF dev BAR2 migration region
hisi_acc_vfio_pci: add new vfio_pci driver for HiSilicon ACC devices
hisi_acc_qm: Move VF PCI device IDs to common header
crypto: hisilicon/qm: Move few definitions to common header
crypto: hisilicon/qm: Move the QM header to include/linux
vfio/mlx5: Fix to not use 0 as NULL pointer
PCI/IOV: Fix wrong kernel-doc identifier
vfio/mlx5: Use its own PCI reset_done error handler
vfio/pci: Expose vfio_pci_core_aer_err_detected()
vfio/mlx5: Implement vfio_pci driver for mlx5 devices
vfio/mlx5: Expose migration commands over mlx5 device
vfio: Remove migration protocol v1 documentation
vfio: Extend the device migration protocol with RUNNING_P2P
vfio: Define device migration protocol v2
...
To avoid repeatedly obtaining 'qm' from 'filp', parameter passing of
debugfs function directly use 'qm' instead of 'filp'.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The accelerator devices support runtime PM, when device is in suspended, an
exception will occur if reading registers. Therefore, this patch uses
'debugfs_create_file' instead of 'debugfs_create_regset32' to create
debugfs file, and then the driver can get the device status before
reading the register.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When the endian configuration of the hardware is abnormal, it will
cause the SEC engine is faulty that reports empty message. And it
will affect the normal function of the hardware. Currently the soft
configuration method can't restore the faulty device. The endian
needs to be configured according to the system properties. So fix it.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Because the algs registration process has added a judgment.
So need to add the judgment for the abnormal exiting process.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Kunpeng930 sec device supports dynamic clock gating. When doing tasks,
the algorithm core is opened, and when idle, the algorithm core is closed.
This patch enables sec dynamic clock gating by writing hardware registers.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The open interface of the sva prefetching function is distinguish the chip
version. But the close interface of the sva prefetching function doesn't
distinguish the chip version. As a result, the sva prefetching close
operation is also performed on Kunpeng920, those registers are important
on Kunpeng920, which eventually leads to abnormal hardware problems. So
need to fix it immediately.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The SEC driver support configure each function's QoS in the Host
for Kunpeng930. The SEC driver needs to configure the maximum shaper
type rate.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Kunpeng930 hardware supports address prefetching to improve performance
before doing tasks in SVA scenario.
This patch enables this function in device initialization by writing
hardware registers. In the process of reset, address prefetching is
disabled to avoid the failure of interaction between accelerator device
and SMMU.
Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Kunpeng930 could be able to close master ooo when NFE occurs, which will
disable memory accessing from device and execute tasks. This ensures that
errors do not spread.
This patch enables the hardware to close master ooo when an error occurs
by writing hardware registers, and ensures that the driver will not drain
qp because the hardware will empty the tasks automatically.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In order to avoid reporting an exception but the error type is not
configured, the driver needs to configure the error type first, and then
enable the error interrupt. Before executing the task, hardware error
initialization is needed so that the hardware can detect the error in time.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Kunpeng930 SEC adds several new hardware error types. This patch enables
the new error types and configures the error types as NFE.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
'err_info' does not support dynamic configuration since it is const type.
Therefore, in order to support new error type later, 'err_info' is changed
to dynamic configuration.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
ctx_q_num is a module parameter set by the user to specify the
number of qp queues required to create a ctx.
When the number of qp queues allocated by PF or VF is less than
the ctx_q_num, an error will be reported when ctx is initialized
in kernel mode, which leads to the problem that the registered
algorithms cannot be used.
Therefore, when PF or VF is initialized, if the number of qp queues
is not enough to create a ctx, the kernel mode cannot be used,
and there is no need to register the kernel mode algorithms.
Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Since the code related to 'CURRENT_QM' debugfs is exactly same in
sec/hpre/zip driver, move 'CURRENT_QM' to qm.c to reduce duplicate code.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move the configuration of the total number of queues 'ctrl_qp_num'
from sec2/hpre/zip to qm.c. And get the total number of queues
from the hardware register for Kunpeng930.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Enable PASID by setting 'sqc' and 'cqc' pasid bits
per queue in Kunpeng 930.
For Kunpeng 920, PASID is effective for all queues once set
in SVA scenarios.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In order to pass kernel CRYPTO test, SEC module parameter
'pf_q_num' needs to be set as greater than 1.
Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When the devices are removed or not existing, the corresponding algorithms
which are registered by 'hisi-zip' driver can't be used.
Move 'hisi_zip_register_to_crypto' from 'hisi_zip_init' to
'hisi_zip_probe'. The algorithms will be registered to crypto only when
there is device bind on the driver. And when the devices are removed,
the algorithms will be unregistered.
In the previous process, the function 'xxx_register_to_crypto' need a lock
and a static variable to judge if the registration is the first time.
Move this action into the function 'hisi_qm_alg_register'. Each device
will call 'hisi_qm_alg_register' to add itself to qm list in probe process
and registering algs when the qm list is empty.
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>
Call trace will appear in the Hisilicon crypto driver unbinding or
disabling SRIOV during task running with TFMs on the corresponding
function.
The log looks like this:
[ 293.908078] Call trace:
[ 293.908080] __queue_work+0x494/0x548
[ 293.908081] queue_work_on+0x84/0xd8
[ 293.908092] qm_irq+0x4c/0xd0 [hisi_qm]
[ 293.908096] __handle_irq_event_percpu+0x74/0x2a0
[ 293.908098] handle_irq_event_percpu+0x40/0x98
[ 293.908099] handle_irq_event+0x4c/0x80
[ 293.908101] handle_fasteoi_irq+0xb0/0x170
[ 293.908102] generic_handle_irq+0x3c/0x58
[ 293.908103] __handle_domain_irq+0x68/0xc0
[ 293.908104] gic_handle_irq+0xb4/0x298
[ 293.908105] el1_irq+0xcc/0x180
[ 293.908107] arch_cpu_idle+0x38/0x228
[ 293.908110] default_idle_call+0x20/0x40
[ 293.908113] do_idle+0x1cc/0x2b8
[ 293.908114] cpu_startup_entry+0x2c/0x30
[ 293.908115] rest_init+0xdc/0xe8
[ 293.908117] arch_call_rest_init+0x14/0x1c
[ 293.908117] start_kernel+0x490/0x4c4
This patch adds a waiting logic as user doing the above two operations
to avoid panic. The two operations will hold on in the driver
remove function until the tasks release all their relative TFMs.
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: Weili Qian <qianweili@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>
Since the drivers such as HPRE/SEC/ZIP do not implement
'pci_driver.shutdow', a RAS will be triggered at OS rebooting or shutting
down as the hardware device is processing request.
The log looks like this:
NOTICE: [NimbusSecNodeType1]:[2372L]This is sec, Base = 0x141800000
NOTICE: [NimbusSecHandle]:[2319L] SecIntSt = 0x3
NOTICE: [NimbusSecHandle]:[2320L] SecQmIntStatus = 0x2
NOTICE: [PrintSecurityType]:[344L] SecurityType is RECOVERABLE!
This patch offers a new API in qm to shutdown devices, and add shutdown
callbacks in ACC driver based on this new API.
So the running devices will be stopped when the OS reboot or shutdown.
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>
Now, there are three reasons of stopping: 'NORMAL', 'SOFT_RESET' and 'FLR'.
In order to keep this, explicitly pass the stop reason as an input
parameter of 'hisi_qm_stop' function.
Fixes: b67202e8ed30("crypto: hisilicon/qm - add state machine for QM")
Signed-off-by: Shukun Tan <tanshukun1@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>
Update debugfs interface parameters, and adjust the
processing logic inside the corresponding function
Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As before, if a SEC queue is at the 'fake busy' status,
the request with a 'fake busy' flag will be sent into hardware
and the sending function returns busy. After the request is
finished, SEC driver's call back will identify the 'fake busy' flag,
and notifies the user that hardware is not busy now by calling
user's call back function.
Now, a request sent into busy hardware will be cached in the
SEC queue's backlog, return '-EBUSY' to user.
After the request being finished, the cached requests will
be processed in the call back function. to notify the
corresponding user that SEC queue can process more requests.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Reviewed-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As stress-ng running SEC engine on the Ubuntu OS,
we found that SEC only supports two threads each with one TFM
based on the default module parameter 'ctx_q_num'.
If running more threads, stress-ng will fail since it cannot
get more TFMs.
In order to fix this, we adjusted the default values
of the module parameters to support more TFMs.
Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Zaibo Xu <xuzaibo@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In order to be compatible with devices of different versions, V1 in the
accelerator driver is now isolated, and other versions are the previous
V2 processing flow.
Signed-off-by: Weili Qian <qianweili@huawei.com>
Signed-off-by: Shukun Tan <tanshukun1@huawei.com>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The hardware device can be configured to report directly through MSI, but
this method will not go through RAS, configure all hardware errors that
should be processed by driver to NFE.
Signed-off-by: Shukun Tan <tanshukun1@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
