Replace a couple of calls to memcpy() with simple assignments in order
to fix the following out-of-bounds warning:
drivers/iommu/intel/svm.c:1198:4: warning: 'memcpy' offset [25, 32] from
the object at 'desc' is out of the bounds of referenced subobject
'qw2' with type 'long long unsigned int' at offset 16 [-Warray-bounds]
The problem is that the original code is trying to copy data into a
couple of struct members adjacent to each other in a single call to
memcpy(). This causes a legitimate compiler warning because memcpy()
overruns the length of &desc.qw2 and &resp.qw2, respectively.
This helps with the ongoing efforts to globally enable -Warray-bounds
and get us closer to being able to tighten the FORTIFY_SOURCE routines
on memcpy().
Link: https://github.com/KSPP/linux/issues/109
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210414201403.GA392764@embeddedor
Link: https://lore.kernel.org/r/20210610020115.1637656-18-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
A debugfs interface /sys/kernel/debug/iommu/intel/dmar_perf_latency is
created to control and show counts of execution time ranges for various
types per DMAR. The interface may help debug any potential performance
issue.
By default, the interface is disabled.
Possible write value of /sys/kernel/debug/iommu/intel/dmar_perf_latency
0 - disable sampling all latency data
1 - enable sampling IOTLB invalidation latency data
2 - enable sampling devTLB invalidation latency data
3 - enable sampling intr entry cache invalidation latency data
4 - enable sampling prq handling latency data
Read /sys/kernel/debug/iommu/intel/dmar_perf_latency gives a snapshot
of sampling result of all enabled monitors.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210520031531.712333-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210610020115.1637656-15-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The execution time of some operations is very performance critical, such
as cache invalidation and PRQ processing time. This adds some common code
to monitor the execution time range of those operations. The interfaces
include enabling/disabling, checking status, updating sampling data and
providing a common string format for users.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210520031531.712333-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210610020115.1637656-14-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Current VT-d implementation supports nested translation only if all
underlying IOMMUs support the nested capability. This is unnecessary
as the upper layer is allowed to create different containers and set
them with different type of iommu backend. The IOMMU driver needs to
guarantee that devices attached to a nested mode iommu_domain should
support nested capabilility.
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210517065701.5078-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210610020115.1637656-6-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The Intel IOMMU driver reports the DMA fault reason in a decimal number
while the VT-d specification uses a hexadecimal one. It's inconvenient
that users need to covert them everytime before consulting the spec.
Let's use hexadecimal number for a DMA fault reason.
The fault message uses 0xffffffff as PASID for DMA requests w/o PASID.
This is confusing. Tweak this by adding "NO_PASID" explicitly.
Reviewed-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210517065425.4953-1-baolu.lu@linux.intel.com
Link: https://lore.kernel.org/r/20210610020115.1637656-4-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The opening comment mark '/**' is used for highlighting the beginning of
kernel-doc comments.
The header for drivers/iommu/intel/pasid.c follows this syntax, but
the content inside does not comply with kernel-doc.
This line was probably not meant for kernel-doc parsing, but is parsed
due to the presence of kernel-doc like comment syntax(i.e, '/**'), which
causes unexpected warnings from kernel-doc:
warning: Function parameter or member 'fmt' not described in 'pr_fmt'
Provide a simple fix by replacing this occurrence with general comment
format, i.e. '/*', to prevent kernel-doc from parsing it.
Signed-off-by: Aditya Srivastava <yashsri421@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20210523143245.19040-1-yashsri421@gmail.com
Link: https://lore.kernel.org/r/20210610020115.1637656-3-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
- Bug fixes for the OCC, SCOM and SBEFIFO drivers
- Performance fix for aspeed fsi master
- Small fixes from the mailing lists
-----BEGIN PGP SIGNATURE-----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=xHRF
-----END PGP SIGNATURE-----
Merge tag 'fsi-for-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/joel/fsi into char-misc-next
Joel writes:
FSI changes for v5.14
- Bug fixes for the OCC, SCOM and SBEFIFO drivers
- Performance fix for aspeed fsi master
- Small fixes from the mailing lists
* tag 'fsi-for-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/joel/fsi:
fsi/sbefifo: Fix reset timeout
fsi/sbefifo: Clean up correct FIFO when receiving reset request from SBE
fsi: master-ast-cf: Remove redundant error printing in fsi_master_acf_probe()
fsi: Aspeed: Reduce poll timeout
fsi: aspeed: convert to devm_platform_ioremap_resource
hwmon: (occ) Print response status in first poll error message
hwmon: (occ) Start sequence number at one
fsi: occ: Log error for checksum failure
fsi: occ: Don't accept response from un-initialized OCC
fsi: scom: Remove retries
fsi: scom: Reset the FSI2PIB engine for any error
fsi: aspeed: Emit fewer barriers in opb operations
fsi: core: Fix return of error values on failures
fsi: Add missing MODULE_DEVICE_TABLE
If access_ok() or fpregs_soft_set() fails in __fpu__restore_sig() then the
function just returns but does not clear the FPU state as it does for all
other fatal failures.
Clear the FPU state for these failures as well.
Fixes: 72a671ced6 ("x86, fpu: Unify signal handling code paths for x86 and x86_64 kernels")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/87mtryyhhz.ffs@nanos.tec.linutronix.de
If the VMM's (Qemu) memory backend is backed up by memfd + Hugepages
(hugetlbfs and not THP), we have to first find the hugepage(s) where
the Guest allocations are located and then extract the regular 4k
sized subpages from them.
v2: Ensure that the subpage and hugepage offsets are calculated correctly
when the range of subpage allocations cuts across multiple hugepages.
v3: Instead of repeatedly looking up the hugepage for each subpage,
only do it when the subpage allocation crosses over into a different
hugepage. (suggested by Gerd and DW)
v4: Fix the following warning identified by checkpatch:
CHECK:OPEN_ENDED_LINE: Lines should not end with a '('
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
Signed-off-by: Dongwon Kim <dongwon.kim@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20210609182915.592743-1-vivek.kasireddy@intel.com
[ kraxel: one more checkpatch format tweak ]
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Added RHBA and RPA attributes type and length.
As per FC_GC_7 document section "Table 400 – Attribute Entry Types and
associated Values" ASCII type attributes length can be vary from "4 to 256
byte". If we keep all RHBA ASCII attributes length 256 then total length
is going upto 2750, which is far more than 2048 (max frame size).
In libfc we do have logic to split FCP commands but not for CT commands.
Practically all version/names get covered with in 64 bytes except OS name,
for that we need 128 bytes. Hence length of all RBHA ASCII attributes
is reduced to 64 bytes and 128 bytes in case of OS name.
RPA attributes total length is within frame size.
Link: https://lore.kernel.org/r/20210603121623.10084-6-jhasan@marvell.com
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add all the attributes for FDMI.
Fall back mechanism is added in between FDMI V2 and FDMI V1 attributes. In
case FDMI get fails for V2 attributes we fall back to V1 attributes.
Link: https://lore.kernel.org/r/20210603121623.10084-5-jhasan@marvell.com
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add all attributes for RHBA and RPA registration.
Fallback mechanism is added between RBHA V2 and RHBA V1 attributes. In case
RHBA get fails for V2 attributes we fall back to V1 attribute registration.
Link: https://lore.kernel.org/r/20210603121623.10084-4-jhasan@marvell.com
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Initialize RHBA and RPA attributes.
Link: https://lore.kernel.org/r/20210603121623.10084-2-jhasan@marvell.com
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Incorrect condition check was leading to data corruption.
Link: https://lore.kernel.org/r/20210603101404.7841-3-jhasan@marvell.com
Fixes: 8fd9efca86 ("scsi: libfc: Work around -Warray-bounds warning")
CC: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
As per the FC-GS-5 specification, attribute lengths of node_name and
manufacturer should in range of "4 to 64 Bytes" only.
Link: https://lore.kernel.org/r/20210603101404.7841-2-jhasan@marvell.com
Fixes: e721eb0616 ("scsi: scsi_transport_fc: Match HBA Attribute Length with HBAAPI V2.0 definitions")
CC: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This fixes 32-bit arm build due to lack of 64-bit divides.
Fixes: cb1c81467a ("drm/ttm: flip the switch for driver allocated resources v2")
Link: https://patchwork.freedesktop.org/patch/438442/
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
If an internal task abort timeout occurs, the controller has developed a
fault, and needs to be reset to be recovered.
When this occurs during error handling, the current policy is to allow
error handling to continue, and the inevitable nexus ha reset will handle
the required reset.
However various steps of error handling need to taken before this happens.
These also involve some level of HW interaction, which will also fail with
various timeouts.
Speed up this process by recording a HW fault bit for an internal abort
timeout - when this is set, just automatically error any HW interaction,
and essentially go straight to clear nexus ha (to reset the controller).
Link: https://lore.kernel.org/r/1623058179-80434-6-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If an internal task abort timeout occurs, the controller has developed a
fault, and needs to be reset to be recovered. However if a timeout occurs
during SCSI error handling, issuing a controller reset immediately may
conflict with the error handling.
To handle internal abort in these two scenarios, only queue the reset when
not in an error handling function. In the case of a timeout during error
handling, do nothing and rely on the inevitable ha nexus reset to reset the
controller.
Link: https://lore.kernel.org/r/1623058179-80434-5-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
For a clear nexus reset operation, the I_T nexus resets are executed
serially for each device. For devices attached through an expander, this
may take 2s per device; so, in total, could take a long time.
Reduce the total time by running the I_T nexus resets in parallel through
async operations.
Link: https://lore.kernel.org/r/1623058179-80434-3-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If an OOB event is received but the phy still fails to come up, a link
reset will be issued repeatedly at an interval of 20s until the phy comes
up.
Set a limit for link reset issue retries to avoid printing the timeout
message endlessly.
Link: https://lore.kernel.org/r/1623058179-80434-2-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
qedi_clear_session_ctx() could race with the in-kernel or userspace driven
recovery/removal and we could access a NULL conn or do a double free.
We should be using iscsi_host_remove() to start the removal process from
the driver. It will start the in-kernel recovery and notify userspace that
the driver's scsi_hosts are being removed. iscsid will then drive the
session removal like is done when the logout command is run. When the
sessions are removed, iscsi_host_remove() will return so qedi can finish
knowing there are no running sessions and no new sessions will be allowed.
This also fixes an issue where we check for a NULL conn after already
accessing it introduced in commit 27e986289e ("scsi: iscsi: Drop suspend
calls from ep_disconnect") by just removing the function completely.
Link: https://lore.kernel.org/r/20210609192709.5094-1-michael.christie@oracle.com
Fixes: 27e986289e ("scsi: iscsi: Drop suspend calls from ep_disconnect")
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The pci_alloc_irq_vectors_affinity() function returns negative error codes
or it returns a number between the minimum vectors (1 in this case) and
max_vectors. It won't return zero. Because "i" is a u16 then the error
handling won't work. And also if it did work the error code was not set.
Really "max_vectors" can be an int as well because we're doing a min_t() on
int type. The other change is that it's better to remove unnecessary
initialization so that static checkers can warn us if there are ever
uninitialized variable bugs introduced in the future.
I changed the error code from -1 (-EPERM) if the kmalloc() failed to
-ENOMEM. And on success path I changed it from "return retval;" to "return
0;" which shouldn't affect the compiled code but makes it more readable.
Link: https://lore.kernel.org/r/YMCJcnmSI4kOIyv/@mwanda
Fixes: 824a156633 ("scsi: mpi3mr: Base driver code")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The "mrioc->intr_info" pointer can't be NULL, but if it could then the
second iteration through the loop would Oops. Let's delete the confusing
and impossible NULL check.
Link: https://lore.kernel.org/r/YMCJKgykDYtyvY44@mwanda
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fix a double free, scsi_tgt_priv_data will be freed in
mpi3mr_target_destroy() so remove the kfree() from mpi3mr_target_alloc().
I've also removed few unneeded initialisations.
Link: https://lore.kernel.org/r/20210608145712.16386-1-thenzl@redhat.com
Acked-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In ufshcd_exec_dev_cmd(), if error happens before lrpb is initialized, then
we should bail out instead of letting trace record the error.
Link: https://lore.kernel.org/r/1623227044-22635-1-git-send-email-cang@codeaurora.org
Fixes: a45f937110 ("scsi: ufs: Optimize host lock on transfer requests send/compl paths")
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Drop disabling of gfxoff during VCN use. This allows gfxoff
to kick in and potentially save power if the user is not using
gfx for color space conversion or scaling.
VCN1.0 had a bug which prevented it from working properly with
gfxoff, so we disabled it while using VCN. That said, most apps
today use gfx for scaling and color space conversion rather than
overlay planes so it was generally in use anyway. This was fixed
on VCN2+, but since we mostly use gfx for color space conversion
and scaling and rapidly powering up/down gfx can negate the
advantages of gfxoff, we left gfxoff disabled. As more
applications use overlay planes for color space conversion
and scaling, this starts to be a win, so go ahead and leave
gfxoff enabled.
Note that VCN1.0 uses vcn_v1_0_idle_work_handler() and
vcn_v1_0_ring_begin_use() so they are not affected by this
patch.
Reviewed-by: James Zhu <James.Zhu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
With support for pvid vlans in mlx5 bridge it is possible to have rules in
untagged flow group when vlan filtering is enabled. However, such rules can
also match tagged packets that didn't match anything in tagged flow group.
Filter such packets by introducing additional flow group between tagged and
untagged groups. When filtering is enabled on the bridge create additional
flow in vlan filtering flow group and matches tagged packets with specified
source MAC address and redirects them to new "skip" table. The skip table
is new lowest-level empty table that is used to skip all further processing
on packet in bridge priority.
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Implement support for pushing vlan header into untagged packet on ingress
of port that has pvid configured and support for popping vlan on egress of
port that has the matching vlan configured as untagged. To support such
configurations packet reformat contexts of {INSERT|REMOVE}_HEADER types are
created per such vlan and saved to struct mlx5_esw_bridge_vlan which allows
all FDB entries on particular vlan to share single packet reformat
instance. When initializing FDB entries with pvid or untagged vlan type set
its mlx5_flow_act->pkt_reformat action accordingly.
Flush all flows when removing vlan from port. This is necessary because
even though software bridge removes all FDB entries before removing their
vlan, mlx5 bridge implementation deletes their corresponding flow entries
from hardware in asynchronous workqueue task, which will cause firmware
error if vlan packet reformat context is deleted before all flows that
point to it.
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Add support for FDB vlan-tagged entries. Extend ingress and egress flow
tables with flow groups to match packet vlan tag. Modify the flow creation
code to include vlan tag, if vlan is configured on port and vlan
configuration is supported for offload.
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Establish all the necessary infrastructure for implementing vlan matching
and vlan push/pop in following patches:
- Add new per-vport struct mlx5_esw_bridge_port that is used to store
metadata for all port vlans. Initialize and cleanup the instance of the
structure when port representor is linked/unliked to bridge. Use xarray to
allow quick vport metadata lookup by vport number.
- Add new per-port-vlan struct mlx5_esw_bridge_vlan that is used to store
vlan-specific data (vid, flags). Handle SWITCHDEV_PORT_OBJ_{ADD|DEL}
switchdev blocking event for SWITCHDEV_OBJ_ID_PORT_VLAN object by
creating/deleting the vlan structure and saving it in per-vport xarray for
quick lookup.
- Implement support for SWITCHDEV_ATTR_ID_BRIDGE_VLAN_FILTERING object
attribute that is used to toggle vlan filtering. Remove all FDB entries
from hardware when vlan filtering state is changed.
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Dynamic FDB entries require capability to age out unused entries. Such
entries are either aged out by kernel software bridge implementation or by
hardware switch that offloaded them (and notified the kernel to mark them
as SWITCHDEV_FDB_ADD_TO_BRIDGE). Leaving ageing to kernel bridge would
result it deleting offloaded dynamic FDB entries every ageing_time period
due to packets being processed by hardware and, consecutively, 'used'
timestamp for FDB entry not being updated. However, since hardware doesn't
support ageing, software solution inside the driver is required.
In order to emulate hardware ageing in driver, extend bridge FDB ingress
flows with counter and create delayed br_offloads->update_work task on
bridge offloads workqueue. Run the task every second, update 'used'
timestamp in software bridge dynamic entry by sending
SWITCHDEV_FDB_ADD_TO_BRIDGE for the entry, if it flow hardware counter
lastuse field was changed since last update. If lastuse wasn't changed for
ageing_time period, then delete the FDB entry and notify kernel bridge by
sending SWITCHDEV_FDB_DEL_TO_BRIDGE notification.
Register blocking switchdev notifier callback and handle attribute set
SWITCHDEV_ATTR_ID_BRIDGE_AGEING_TIME event to allow user to dynamically
configure bridge FDB entry ageing timeout. Save the value per-bridge in
struct mlx5_esw_bridge. Silently ignore
SWITCHDEV_ATTR_ID_PORT_{PRE_}BRIDGE_FLAGS switchdev event since mlx5 bridge
implementation relies on software bridge for implementing necessary
behavior for all of these flags.
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Hardware supported by mlx5 driver doesn't provide learning and requires the
driver to emulate all switch-like behavior in software. As such, all
packets by default go through miss path, appear on representor and get to
software bridge, if it is the upper device of the representor. This causes
bridge to process packet in software, learn the MAC address to FDB and send
SWITCHDEV_FDB_ADD_TO_DEVICE event to all subscribers.
In order to offload FDB entries in mlx5, register switchdev notifier
callback and implement support for both 'added_by_user' and dynamic FDB
entry SWITCHDEV_FDB_ADD_TO_DEVICE events asynchronously using new
mlx5_esw_bridge_offloads->wq ordered workqueue. In workqueue callback
offload the ingress rule (matching FDB entry MAC as packet source MAC) and
egress table rule (matching FDB entry MAC as destination MAC). For ingress
table rule also match source vport to ensure that only traffic coming from
expected bridge port is matched by offloaded rule. Save all the relevant
FDB entry data in struct mlx5_esw_bridge_fdb_entry instance and insert the
instance in new mlx5_esw_bridge->fdb_list list (for traversing all entries
by software ageing implementation in following patch) and in new
mlx5_esw_bridge->fdb_ht hash table for fast retrieval. Notify the bridge
that FDB entry has been offloaded by sending SWITCHDEV_FDB_OFFLOADED
notification.
Delete FDB entry on reception of SWITCHDEV_FDB_DEL_TO_DEVICE event.
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Create new files bridge.{c|h} in en/rep directory that implement bridge
interaction with representor netdevices and handle required
events/notifications, bridge.{c|h} in esw directory that implement all
necessary eswitch offloading infrastructure and works on vport/eswitch
level. Provide new kconfig MLX5_BRIDGE which is automatically selected when
both kernel bridge and mlx5 eswitch configs are enabled.
Provide basic infrastructure for bridge offloads:
- struct mlx5_esw_bridge_offloads - per-eswitch bridge offload structure
that encapsulates generic bridge-offloads data (notifier blocks, ingress
flow table/group, etc.) that is created/deleted on enable/disable eswitch
offloads.
- struct mlx5_esw_bridge - per-bridge structure that encapsulates
per-bridge data (reference counter, FDB, egress flow table/group, etc.)
that is created when first eswitch represetor is attached to new bridge and
deleted when last representor is removed from the bridge as a result of
NETDEV_CHANGEUPPER event.
The bridge tables are created with new priority FDB_BR_OFFLOAD in FDB
namespace. The new priority is between tc-miss and slow path priorities.
Priority consist of two levels: the ingress table that is global per
eswitch and matches incoming packets by src_mac/vid and redirects them to
next level (egress table) that is chosen according to ingress port bridge
membership and matches on dst_mac/vid in order to redirect packet to vport
according to the following diagram:
+
|
+---------v----------+
| |
| FDB_TC_OFFLOAD |
| |
+---------+----------+
|
|
+---------v----------+
| |
| FDB_FT_OFFLOAD |
| |
+---------+----------+
|
|
+---------v----------+
| |
| FDB_TC_MISS |
| |
+---------+----------+
|
+--------------------------------------+
| | |
| +------+ |
| | |
| +------v--------+ FDB_BR_OFFLOAD |
| | INGRESS_TABLE | |
| +------+---+----+ |
| | | match |
| | +---------+ |
| | | | +-------+
| | +-------v-------+ match | | |
| | | EGRESS_TABLE +------------> vport |
| | +-------+-------+ | | |
| | | | +-------+
| | miss | |
| +------+------+ |
| | |
+--------------------------------------+
|
|
+---------v----------+
| |
| FDB_SLOW_PATH |
| |
+---------+----------+
|
v
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>