- 01 Dec, 2022 23 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxJakub Kicinski authored
Saeed Mahameed says: ==================== mlx5-updates-2022-11-29 Misc update for mlx5 driver 1) Various trivial cleanups 2) Maor Dickman, Adds support for trap offload with additional actions 3) From Tariq, UMR (device memory registrations) cleanups, UMR WQE must be aligned to 64B per device spec, (not a bug fix). * tag 'mlx5-updates-2022-11-29' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux: net/mlx5e: Support devlink reload of IPsec core net/mlx5e: TC, Add offload support for trap with additional actions net/mlx5e: Do early return when setup vports dests for slow path flow net/mlx5: Remove redundant check net/mlx5e: Delete always true DMA check net/mlx5e: Don't access directly DMA device pointer net/mlx5e: Don't use termination table when redundant net/mlx5: Fix orthography errors in documentation net/mlx5: Use generic definition for UMR KLM alignment net/mlx5: Generalize name of UMR alignment definition net/mlx5: Remove unused UMR MTT definitions net/mlx5e: Add padding when needed in UMR WQEs net/mlx5: Remove unused ctx variables net/mlx5e: Replace zero-length arrays with DECLARE_FLEX_ARRAY() helper net/mlx5e: Remove unneeded io-mapping.h #include ==================== Link: https://lore.kernel.org/r/20221130051152.479480-1-saeed@kernel.orgSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Xiaolei Wang authored
If the external phy used by current mac interface is managed by another mac interface, it means that this network port cannot work independently, especially when the system suspends and resumes, the following trace may appear, so we should create a device link between phy dev and mac dev. WARNING: CPU: 0 PID: 24 at drivers/net/phy/phy.c:983 phy_error+0x20/0x68 Modules linked in: CPU: 0 PID: 24 Comm: kworker/0:2 Not tainted 6.1.0-rc3-00011-g5aaef24b-dirty #34 Hardware name: Freescale i.MX6 SoloX (Device Tree) Workqueue: events_power_efficient phy_state_machine unwind_backtrace from show_stack+0x10/0x14 show_stack from dump_stack_lvl+0x68/0x90 dump_stack_lvl from __warn+0xb4/0x24c __warn from warn_slowpath_fmt+0x5c/0xd8 warn_slowpath_fmt from phy_error+0x20/0x68 phy_error from phy_state_machine+0x22c/0x23c phy_state_machine from process_one_work+0x288/0x744 process_one_work from worker_thread+0x3c/0x500 worker_thread from kthread+0xf0/0x114 kthread from ret_from_fork+0x14/0x28 Exception stack(0xf0951fb0 to 0xf0951ff8) Signed-off-by:
Xiaolei Wang <xiaolei.wang@windriver.com> Tested-by:
Florian Fainelli <f.fainelli@gmail.com> Reviewed-by:
-
Jakub Kicinski authored
Vincent Mailhol says: ==================== net: devlink: return the driver name in devlink_nl_info_fill The driver name is available in device_driver::name. Right now, drivers still have to report this piece of information themselves in their devlink_ops::info_get callback function. The goal of this series is to have the devlink core to report this information instead of the drivers. The first patch fulfills the actual goal of this series: modify devlink core to report the driver name and clean-up all drivers. Both have to be done in an atomic change to avoid attribute duplication. This same patch also removes the devlink_info_driver_name_put() function to prevent future drivers from reporting the driver name themselves. The second patch allows the core to call devlink_nl_info_fill() even if the devlink_ops::info_get() callback is NULL. This leads to the third and final patch which cleans up the drivers which have an empty info_get(). ==================== Link: https://lore.kernel.org/r/20221129095140.3913303-1-mailhol.vincent@wanadoo.frSigned-off-by:
-
Vincent Mailhol authored
devlink_ops::info_get() is now optional and devlink will continue to report information even if that callback gets removed. Remove all the empty devlink_ops::info_get() callbacks from the drivers. Signed-off-by:
Vincent Mailhol <mailhol.vincent@wanadoo.fr> Reviewed-by:
Jacob Keller <jacob.e.keller@intel.com> Reviewed-by:
Jiri Pirko <jiri@nvidia.com> Signed-off-by:
-
Vincent Mailhol authored
Some drivers only reported the driver name in their devlink_ops::info_get() callback. Now that the core provides this information, the callback became empty. For such drivers, just removing the callback would prevent the core from executing devlink_nl_info_fill() meaning that "devlink dev info" would not return anything. Make the callback function optional by executing devlink_nl_info_fill() even if devlink_ops::info_get() is NULL. N.B.: the drivers with devlink support which previously did not implement devlink_ops::info_get() will now also be able to report the driver name. Signed-off-by:
-
Vincent Mailhol authored
The driver name is available in device_driver::name. Right now, drivers still have to report this piece of information themselves in their devlink_ops::info_get callback function. In order to factorize code, make devlink_nl_info_fill() add the driver name attribute. Now that the core sets the driver name attribute, drivers are not supposed to call devlink_info_driver_name_put() anymore. Remove devlink_info_driver_name_put() and clean-up all the drivers using this function in their callback. Signed-off-by:
-
Luca Weiss authored
The NQ310 is another NFC chip from NXP, document the compatible in the bindings. Signed-off-by:
Luca Weiss <luca@z3ntu.xyz> Acked-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Link: https://lore.kernel.org/r/20221128173744.833018-1-luca@z3ntu.xyzSigned-off-by:
-
Jakub Kicinski authored
Jacob Keller says: ==================== support direct read from region A long time ago when initially implementing devlink regions in ice I proposed the ability to allow reading from a region without taking a snapshot [1]. I eventually dropped this work from the original series due to size. Then I eventually lost track of submitting this follow up. This can be useful when interacting with some region that has some definitive "contents" from which snapshots are made. For example the ice driver has regions representing the contents of the device flash. If userspace wants to read the contents today, it must first take a snapshot and then read from that snapshot. This makes sense if you want to read a large portion of data or you want to be sure reads are consistently from the same recording of the flash. However if user space only wants to read a small chunk, it must first generate a snapshot of the entire contents, perform a read from the snapshot, and then delete the snapshot after reading. For such a use case, a direct read from the region makes more sense. This can be achieved by allowing the devlink region read command to work without a snapshot. Instead the portion to be read can be forwarded directly to the driver via a new .read callback. This avoids the need to read the entire region contents into memory first and avoids the software overhead of creating a snapshot and then deleting it. This series implements such behavior and hooks up the ice NVM and shadow RAM regions to allow it. [1] https://lore.kernel.org/netdev/20200130225913.1671982-1-jacob.e.keller@intel.com/ ==================== Link: https://lore.kernel.org/r/20221128203647.1198669-1-jacob.e.keller@intel.comSigned-off-by:
-
Jacob Keller authored
Implement the .read handler for the NVM and Shadow RAM regions. This enables user space to read a small chunk of the flash without needing the overhead of creating a full snapshot. Update the documentation for ice to detail which regions have direct read support. Signed-off-by:
-
Jacob Keller authored
78ad87da ("ice: devlink: add shadow-ram region to snapshot Shadow RAM") added support for the 'shadow-ram' devlink region, but did not document it in the ice devlink documentation. Fix this. Signed-off-by:
-
Jacob Keller authored
The ice driver supports a region for both the flat NVM contents as well as the Shadow RAM which is a layer built on top of the flash during device initialization. These regions use an almost identical read function, except that the NVM needs to set the direct flag when reading, while Shadow RAM needs to read without the direct flag set. They each call ice_read_flat_nvm with the only difference being whether to set the direct flash flag. The NVM region read function also was fixed to read the NVM in blocks to avoid a situation where the firmware reclaims the lock due to taking too long. Note that the region snapshot function takes the ops pointer so the function can easily determine which region to read. Make use of this and re-use the NVM snapshot function for both the NVM and Shadow RAM regions. This makes Shadow RAM benefit from the same block approach as the NVM region. It also reduces code in the ice driver. Signed-off-by:
-
Jacob Keller authored
To read from a region, user space must currently request a new snapshot of the region and then read from that snapshot. This can sometimes be overkill if user space only reads a tiny portion. They first create the snapshot, then request a read, then destroy the snapshot. For regions which have a single underlying "contents", it makes sense to allow supporting direct reading of the region data. Extend the DEVLINK_CMD_REGION_READ to allow direct reading from a region if requested via the new DEVLINK_ATTR_REGION_DIRECT. If this attribute is set, then perform a direct read instead of using a snapshot. Direct read is mutually exclusive with DEVLINK_ATTR_REGION_SNAPSHOT_ID, and care is taken to ensure that we reject commands which provide incorrect attributes. Regions must enable support for direct read by implementing the .read() callback function. If a region does not support such direct reads, a suitable extended error message is reported. Signed-off-by:
-
Jacob Keller authored
The devlink_nl_region_read_snapshot_fill is used to copy the contents of a snapshot into a message for reporting to userspace via the DEVLINK_CMG_REGION_READ netlink message. A future change is going to add support for directly reading from a region. Almost all of the logic for this new capability is identical. To help reduce code duplication and make this logic more generic, refactor the function to take a cb and cb_priv pointer for doing the actual copy. Add a devlink_region_snapshot_fill implementation that will simply copy the relevant chunk of the region. This does require allocating some storage for the chunk as opposed to simply passing the correct address forward to the devlink_nl_cmg_region_read_chunk_fill function. A future change to implement support for directly reading from a region without a snapshot will provide a separate implementation that calls the newly added devlink region operation. Signed-off-by:
-
Jacob Keller authored
The devlink parameter of the devlink_nl_cmd_region_read_chunk_fill function is not used. Remove it, to simplify the function signature. Once removed, it is also obvious that the devlink parameter is not necessary for the devlink_nl_region_read_snapshot_fill either. Signed-off-by:
-
Jacob Keller authored
The snapshot pointer is obtained inside of the function devlink_nl_region_read_snapshot_fill. Simplify this function by locating the snapshot upfront in devlink_nl_cmd_region_read_dumpit instead. This aligns with how other netlink attributes are handled, and allows us to exit slightly earlier if an invalid snapshot ID is provided. It also allows us to pass the snapshot pointer directly to the devlink_nl_region_read_snapshot_fill, and remove the now unused attrs parameter. Signed-off-by:
-
Jacob Keller authored
Report extended error details in the devlink_nl_cmd_region_read_dumpit() function, by using the extack structure from the netlink_callback. Signed-off-by:
-
Jacob Keller authored
The calculation for the data_size in the devlink_nl_read_snapshot_fill function uses an if statement that is better expressed using the min_t macro. Signed-off-by:
-
Horatiu Vultur authored
Currently whenever a new rule id is generated, it picks up the next number bigger than previous id. So it would always be 1, 2, 3, etc. When the rule with id 1 will be deleted and a new rule will be added, it will have the id 4 and not id 1. In theory this can be a problem if at some point a rule will be added and removed ~0 times. Then no more rules can be added because there are no more ids. Change this such that when a new rule is added, search for an empty rule id starting with value of 1 as value 0 is reserved. Fixes: c9da1ac1 ("net: microchip: sparx5: Adding initial tc flower support for VCAP API") Signed-off-by:
Horatiu Vultur <horatiu.vultur@microchip.com> Link: https://lore.kernel.org/r/20221128142959.8325-1-horatiu.vultur@microchip.comSigned-off-by:
-
Christophe JAILLET authored
We know that table_size = table->mem_table.depth * table->mem_table.ways, so use it instead, it is less verbose. Signed-off-by:
Christophe JAILLET <christophe.jaillet@wanadoo.fr> Link: https://lore.kernel.org/r/5230dabe27f48948a9fd0f50a62e2437b65e6a6e.1669378798.git.christophe.jaillet@wanadoo.frSigned-off-by:
-
Christophe JAILLET authored
This allocation is really spurious. The size of the bitmap is 'tot_ids' and it is used as such in the driver. So we could expect something like: table->id_bmap = devm_kcalloc(rvu->dev, BITS_TO_LONGS(table->tot_ids), sizeof(long), GFP_KERNEL); However, when the bitmap is allocated, we allocate: BITS_TO_LONGS(table->tot_ids) * table->tot_ids ~= table->tot_ids / 32 * table->tot_ids ~= table->tot_ids^2 / 32 It is proportional to the square of 'table->tot_ids' which seems to potentially be big. Allocate the expected amount of memory, and switch to the bitmap API to have it more straightforward. Signed-off-by:
-
Christophe JAILLET authored
Use devm_bitmap_zalloc() instead of hand-writing it. This also makes the comment "Allocate bitmap for 32 entry mcam" more explicit because now 32 is really used in the allocation function, instead of an obscure 'sizeof(long)'. Signed-off-by:
-
Christophe JAILLET authored
Use kzalloc() instead of kmalloc()/memset(). Signed-off-by:
-
Christophe JAILLET authored
When this error message is displayed, we know that the all the bits in the bitmap are set. So, bitmap_weight() will return the number of bits of the bitmap, which is 'table->tot_ids'. It is unlikely that a bit will be cleared between mutex_unlock() and dev_err(), but, in order to simplify the code and avoid this possibility, just take 'table->tot_ids'. Signed-off-by:
-
- 30 Nov, 2022 17 commits
-
-
Willem de Bruijn authored
Test NIC hardware checksum offload: - Rx + Tx - IPv4 + IPv6 - TCP + UDP Optional features: - zero checksum 0xFFFF - checksum disable 0x0000 - transport encap headers - randomization See file header for detailed comments. Expected results differ depending on NIC features: - CHECKSUM_UNNECESSARY vs CHECKSUM_COMPLETE - NETIF_F_HW_CSUM (csum_start/csum_off) vs NETIF_F_IP(V6)_CSUM Signed-off-by:
Willem de Bruijn <willemb@google.com> Link: https://lore.kernel.org/r/20221128140210.553391-1-willemdebruijn.kernel@gmail.comSigned-off-by:
-
Leon Romanovsky authored
Change IPsec initialization flow to allow future creation of hardware resources that should be released and allocated during devlink reload operation. As part of that change, update function signature to be void as no callers are actually interested in it. Reviewed-by:
Saeed Mahameed <saeedm@nvidia.com> Signed-off-by:
Leon Romanovsky <leonro@nvidia.com> Signed-off-by:
-
Maor Dickman authored
TC trap action offload is currently supported only when trap is the sole action in the flow. This patch remove this limitation by changing trap action offload to not use MLX5_ATTR_FLAG_SLOW_PATH flag and instead set the flow destination table explicitly to be the slow table. This will allow offload of the additional actions. TC flow example: tc filter add dev $REP2 protocol ip prio 2 root \ flower skip_sw dst_mac $mac0 \ action mirred egress redirect dev $REP3 \ action pedit ex munge eth dst set $mac2 pipe \ action trap Signed-off-by:
Maor Dickman <maord@nvidia.com> Reviewed-by:
Raed Salem <raeds@nvidia.com> Reviewed-by:
Roi Dayan <roid@nvidia.com> Signed-off-by:
-
Roi Dayan authored
Adding flow flag cases in setup vport dests before the slow path case is incorrect as the slow path should take precedence. Current code doesn't show this importance so make the slow path case return early and separate from the other cases and remove the redundant comparison of it in the sample case. Signed-off-by:
Chris Mi <cmi@nvidia.com> Reviewed-by:
Oz Shlomo <ozsh@nvidia.com> Signed-off-by:
-
Leon Romanovsky authored
If ASO failed in creation, it won't be called to destroy either. The kernel coding pattern is to make sure that callers are calling to destroy only for valid objects. Reviewed-by:
-
Leon Romanovsky authored
DMA address always exists for MACsec ASO object. Reviewed-by:
-
Leon Romanovsky authored
Use specialized helper to fetch DMA device pointer. Reviewed-by:
-
Roi Dayan authored
Current code used termination table for each vport destination while it's only required for hairpin, i.e. uplink to uplink, or when vlan push on rx action being used. Fix to skip using termination table for vport destinations that do not require it. Signed-off-by:
-
Rahul Rameshbabu authored
Improve general readability of the device driver documentation. Signed-off-by:
Rahul Rameshbabu <rrameshbabu@nvidia.com> Reviewed-by:
Tariq Toukan <tariqt@nvidia.com> Reviewed-by:
Gal Pressman <gal@nvidia.com> Signed-off-by:
-
Tariq Toukan authored
MLX5_UMR_KLM_ALIGNMENT is in units of number of entries, while MLX5_UMR_MTT_ALIGNMENT (generalized and renamed to MLX5_UMR_FLEX_ALIGNMENT) is in byte units. This is misleading and confusing. Replace this KLM definition with one based on the generic definition. Signed-off-by:
-
Tariq Toukan authored
Per the device spec, MLX5_UMR_MTT_ALIGNMENT is good not only for UMR MTT entries, but for all other entries as well, like KLMs and KSMs. Signed-off-by:
-
Tariq Toukan authored
Defines MLX5_UMR_MTT_MASK and MLX5_UMR_MTT_MIN_CHUNK_SIZE are not in use. Remove them. Signed-off-by:
-
Tariq Toukan authored
Per the device spec, MTTs/KLMs list in a UMR WQE must be aligned to 64B. Per our SW design, the MTT/KLMs list would need alignment only if it's too small, for example on PPC when PAGE_SIZE is 64KB, and only 4 pages are needed to cover a MPWQE of size 256KB. Padding, if needed, is taken into account when calculating the UMR WQE fields (ds_cnt and xlt_octowords), however no entries are provided, instead garbage is passed. No real harm though, as these parts act as gaps between the RX MPWQEs and not used by any of them. Hence, in practice, device does not try to write any incoming packet to them. Still, prefer providing clean padding marking the end of the list, and do not map garbage into the RQ memory region. Signed-off-by:
-
Petr Pavlu authored
Remove mlx5_priv.ctx_list and ctx_lock which are no longer used after commit 601c10c8 ("net/mlx5: Delete custom device management logic"). Signed-off-by:
Petr Pavlu <petr.pavlu@suse.com> Signed-off-by:
-
Gustavo A. R. Silva authored
Zero-length arrays are deprecated and we are moving towards adopting C99 flexible-array members, instead. So, replace zero-length arrays declarations in anonymous union with the new DECLARE_FLEX_ARRAY() helper macro. This helper allows for flexible-array members in unions. Link: https://github.com/KSPP/linux/issues/193 Link: https://github.com/KSPP/linux/issues/222 Link: https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.htmlSigned-off-by:
Gustavo A. R. Silva <gustavoars@kernel.org> Reviewed-by:
Kees Cook <keescook@chromium.org> Signed-off-by:
-
Christophe JAILLET authored
The mlx5 net files don't use io_mapping functionalities. So there is no point in including <linux/io-mapping.h>. Remove it. Signed-off-by:
Pavan Chebbi <pavan.chebbi@broadcom.com> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-nextJakub Kicinski authored
Steffen Klassert says: ==================== ipsec-next 2022-11-26 1) Remove redundant variable in esp6. From Colin Ian King. 2) Update x->lastused for every packet. It was used only for outgoing mobile IPv6 packets, but showed to be usefull to check if the a SA is still in use in general. From Antony Antony. 3) Remove unused variable in xfrm_byidx_resize. From Leon Romanovsky. 4) Finalize extack support for xfrm. From Sabrina Dubroca. * 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next: xfrm: add extack to xfrm_set_spdinfo xfrm: add extack to xfrm_alloc_userspi xfrm: add extack to xfrm_do_migrate xfrm: add extack to xfrm_new_ae and xfrm_replay_verify_len xfrm: add extack to xfrm_del_sa xfrm: add extack to xfrm_add_sa_expire xfrm: a few coding style clean ups xfrm: Remove not-used total variable xfrm: update x->lastused for every packet esp6: remove redundant variable err ==================== Link: https://lore.kernel.org/r/20221126110303.1859238-1-steffen.klassert@secunet.comSigned-off-by:
-
