- 13 Feb, 2023 24 commits
-
-
Shannon Nelson authored
Minor Code cleanup details. Signed-off-by:
Shannon Nelson <shannon.nelson@amd.com> Reviewed-by:
Leon Romanovsky <leonro@nvidia.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Shannon Nelson authored
We have the pointer already, don't need to go through the lif struct for it. Signed-off-by:
-
David S. Miller authored
Alex Elder says: ==================== net: ipa: determine GSI register offsets differently This series changes the way GSI register offset are specified, using the "reg" mechanism currently used for IPA registers. A follow-on series will extend this work so fields within GSI registers are also specified this way. The first patch rearranges the GSI register initialization code so it is similar to the way it's done for the IPA registers. The second identifies all the GSI registers in an enumerated type. The third introduces "gsi_reg-v3.1.c" and uses the "reg" code to define one GSI register offset. The second-to-last patch just adds "gsi_reg-v3.5.1.c", because that version introduces a new register not previously defined. All the rest just define the rest of the GSI register offsets using the "reg" mechanism. Note that, to have continued lines align with an open parenthesis, new files created in this series cause some checkpatch warnings. ==================== Signed-off-by: -
Alex Elder authored
Add the remaining GSI register offset definitions. Use gsi_reg() rather than the corresponding GSI_*_OFFSET() macros to get the offsets for these registers, and get rid of the macros. Note that we are now defining information for the HW_PARAM_2 register, and that doesn't appear until IPA v3.5.1. Signed-off-by:
Alex Elder <elder@linaro.org> Signed-off-by:
-
Alex Elder authored
The next patch adds a GSI register field that is only valid starting at IPA v3.5.1. Create "gsi_v3.5.1.c" from "gsi_v3.1.c", changing only the name of the public regs structure it defines. Signed-off-by:
-
Alex Elder authored
Add definitions of the offsets for IRQ-related GSI registers. Use gsi_reg() rather than the corresponding GSI_CNTXT_*_OFFSET() macros to get the offsets for these registers, and get rid of the macros. Signed-off-by:
-
Alex Elder authored
Add definitions of the offsets and strides for registers whose offset depends on an event ring ID, and use gsi_reg() and its returned value to determine offsets for these registers. Get rid of the corresponding GSI_EV_CH_E_*_OFFSET() macros. Signed-off-by:
-
Alex Elder authored
Continue populating with GSI register definitions, adding remaining registers whose offset depends on a channel ID. Use gsi_reg() and reg_n_offset() to determine offsets for those registers, and get rid of the corresponding GSI_CH_C_*_OFFSET() macros. Signed-off-by:
-
Alex Elder authored
Create a new register definition file in the "reg" subdirectory, and begin populating it with GSI register definitions based on IPA version. The GSI registers haven't changed much, so several IPA versions can share the same GSI register definitions. As with IPA registers, an array of pointers indexed by GSI register ID refers to these register definitions, and a new "regs" field in the GSI structure is initialized in gsi_reg_init() to refer to register information based on the IPA version (though for now there's only one). The new function gsi_reg() returns register information for a given GSI register, and the result can be used to look up that register's offset. This patch is meant only to put the infrastructure in place, so only eon register (CH_C_QOS) is defined for each version, and only the offset and stride are defined for that register. Use new function gsi_reg() to look up that register's information to get its offset, This makes the GSI_CH_C_QOS_OFFSET() unnecessary, so get rid of it. Signed-off-by:
-
Alex Elder authored
Create a new gsi_reg_id enumerated type, which identifies each GSI register with a symbolic identifier. Create a function that indicates whether a register ID is valid. Signed-off-by:
-
Alex Elder authored
Create a new source file "gsi_reg.c", and in it, introduce a new function to encapsulate initializing GSI registers, including looking up and I/O mapping their memory. Create gsi_reg_exit() as the inverse of the init function. Signed-off-by:
-
Eric Dumazet authored
If TCA_STAB attribute is malformed, qdisc_get_stab() returns an error, and we end up calling ops->destroy() while ops->init() has not been called yet. While we are at it, call qdisc_put_stab() after ops->destroy(). Fixes: 1f62879e ("net/sched: make stab available before ops->init() call") Reported-by: syzbot+d44d88f1d11e6ca8576b@syzkaller.appspotmail.com Signed-off-by:
Eric Dumazet <edumazet@google.com> Cc: Vladimir Oltean <vladimir.oltean@nxp.com> Cc: Kurt Kanzenbach <kurt@linutronix.de> Reviewed-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by:
-
Horatiu Vultur authored
Add support for PHC and timestamping operations for the lan8841 PHY. PTP 1-step and 2-step modes are supported, over Ethernet and UDP both ipv4 and ipv6. Signed-off-by:
Horatiu Vultur <horatiu.vultur@microchip.com> Signed-off-by:
-
David S. Miller authored
Jiri Pirko says: ==================== devlink: params cleanups and devl_param_driverinit_value_get() fix The primary motivation of this patchset is the patch #6, which fixes an issue introduced by 075935f0 ("devlink: protect devlink param list by instance lock") and reported by Kim Phillips <kim.phillips@amd.com> (https://lore.kernel.org/netdev/719de4f0-76ac-e8b9-38a9-167ae239efc7@amd.com/) and my colleagues doing mlx5 driver regression testing. The basis idea is that devl_param_driverinit_value_get() could be possible to the called without holding devlink intance lock in most of the cases (all existing ones in the current codebase), which would fix some mlx5 flows where the lock is not held. To achieve that, make sure that the param value does not change between reloads with patch #2. Also, convert the param list to xarray which removes the worry about list_head consistency when doing lockless lookup. The rest of the patches are doing some small related cleanup of things that poke me in the eye during the work. --- v1->v2: - a small bug was fixed in patch #2, the rest of the code stays the same so I left review/ack tags attached to them ==================== Signed-off-by:
-
Jiri Pirko authored
Driver calling devl_param_driverinit_value_set() has to hold devlink instance lock while doing that. Put an assertion there. Signed-off-by:
Jiri Pirko <jiri@nvidia.com> Reviewed-by:
Simon Horman <simon.horman@corigine.com> Acked-by:
Jakub Kicinski <kuba@kernel.org> Reviewed-by:
Jacob Keller <jacob.e.keller@intel.com> Signed-off-by:
-
Jiri Pirko authored
If the driver maintains following basic sane behavior, the devl_param_driverinit_value_get() function could be called without holding instance lock: 1) Driver ensures a call to devl_param_driverinit_value_get() cannot race with registering/unregistering the parameter with the same parameter ID. 2) Driver ensures a call to devl_param_driverinit_value_get() cannot race with devl_param_driverinit_value_set() call with the same parameter ID. 3) Driver ensures a call to devl_param_driverinit_value_get() cannot race with reload operation. By the nature of params usage, these requirements should be trivially achievable. If the driver for some off reason is not able to comply, it has to take the devlink->lock while calling devl_param_driverinit_value_get(). Remove the lock assertion and add comment describing the locking requirements. This fixes a splat in mlx5 driver introduced by the commit referenced in the "Fixes" tag. Lore: https://lore.kernel.org/netdev/719de4f0-76ac-e8b9-38a9-167ae239efc7@amd.com/Reported-by:
Kim Phillips <kim.phillips@amd.com> Fixes: 075935f0 ("devlink: protect devlink param list by instance lock") Signed-off-by:
-
Jiri Pirko authored
Loose the linked list for params and use xarray instead. Note that this is required to be eventually possible to call devl_param_driverinit_value_get() without holding instance lock. Signed-off-by:
-
Jiri Pirko authored
As xarray has an iterator helper that allows to start from specified index, use this directly and avoid repeated iteration from 0. Signed-off-by:
-
Jiri Pirko authored
Probably due to copy-paste error, the name of the arg is "init_val" which is misleading, as the pointer is used to point to struct where to store the current value. Rename it to "val" and change the arg comment a bit on the way. Signed-off-by:
-
Jiri Pirko authored
The driverinit param purpose is to serve the driver during init/reload time to provide a value, either default or set by user. Make sure that driver does not read value updated by user before the reload is performed. Hold the new value in a separate struct and switch it during reload. Note that this is required to be eventually possible to call devl_param_driverinit_value_get() without holding instance lock. Signed-off-by:
-
Jiri Pirko authored
No need to treat string params any different comparing to other types. Rely on the struct assign to copy the whole struct, including the string. Signed-off-by:
-
Yu Xiao authored
Add support for the following modes to the nfp driver: NFP_MEDIA_10GBASE_LR NFP_MEDIA_25GBASE_LR NFP_MEDIA_25GBASE_ER These modes are supported by the hardware and, support for them was recently added to firmware. Signed-off-by:
Yu Xiao <yu.xiao@corigine.com> Signed-off-by:
-
Jacob Keller authored
NL_SET_ERR_MSG_MOD inserts the KBUILD_MODNAME and a ':' before the actual extended error message. The devlink feature hasn't been able to be compiled as a module since commit f4b6bcc7 ("net: devlink: turn devlink into a built-in"). Stop using NL_SET_ERR_MSG_MOD, and just use the base NL_SET_ERR_MSG. This aligns the extended error messages better with the NL_SET_ERR_MSG_ATTR messages as well. Signed-off-by:
-
Pietro Borrello authored
rds_rm_zerocopy_callback() uses list_add_tail() with swapped arguments. This links the list head with the new entry, losing the references to the remaining part of the list. Fixes: 9426bbc6 ("rds: use list structure to track information for zerocopy completion notification") Suggested-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
Pietro Borrello <borrello@diag.uniroma1.it> Signed-off-by:
-
- 11 Feb, 2023 16 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queueJakub Kicinski authored
Tony Nguyen says: ==================== Intel Wired LAN Driver Updates 2023-02-09 (i40e) Jan removes i40e_status from the driver; replacing them with standard kernel error codes. Kees Cook replaces 0-length array with flexible array. * '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue: net/i40e: Replace 0-length array with flexible array i40e: use ERR_PTR error print in i40e messages i40e: use int for i40e_status i40e: Remove string printing for i40e_status i40e: Remove unused i40e status codes ==================== Link: https://lore.kernel.org/r/20230209172536.3595838-1-anthony.l.nguyen@intel.comSigned-off-by:
-
Jakub Kicinski authored
Alexandra Winter says: ==================== s390/net: updates 2023-02-06 Just maintenance patches, no functional changes. If you disagree with patch 4, we can leave it out. We prefer scnprintf, but no strong opinion. ==================== Link: https://lore.kernel.org/r/20230209110424.1707501-1-wintera@linux.ibm.comSigned-off-by:
-
Thorsten Winkler authored
This LWN article explains the rationale for this change https: //lwn.net/Articles/69419/ Ie. snprintf() returns what *would* be the resulting length, while scnprintf() returns the actual length. Reported-by:
Jules Irenge <jbi.octave@gmail.com> Reviewed-by:
Alexandra Winter <wintera@linux.ibm.com> Signed-off-by:
Thorsten Winkler <twinkler@linux.ibm.com> Signed-off-by:
-
Thorsten Winkler authored
Following the advice of the Documentation/filesystems/sysfs.rst. All sysfs related show()-functions should only use sysfs_emit() or sysfs_emit_at() when formatting the value to be returned to user space. Reported-by:
Joe Perches <joe@perches.com> Reviewed-by:
-
Thorsten Winkler authored
Use INET6_ADDRSTRLEN constant with size of 48 which be used for char arrays storing ip addresses (for IPv4 and IPv6) Reviewed-by:
-
Alexandra Winter authored
Get rid of multiple smatch warnings, like: warn: inconsistent indenting Reviewed-by:
-
Jakub Kicinski authored
Yoshihiro Shimoda says: ==================== net: renesas: rswitch: Improve TX timestamp accuracy The patch [[123]/4] are minor refacoring for readability. The patch [4/4] is for improving TX timestamp accuracy. To improve the accuracy, it requires refactoring so that this is not a fixed patch. v2: https://lore.kernel.org/all/20230208235721.2336249-1-yoshihiro.shimoda.uh@renesas.com/ v1: https://lore.kernel.org/all/20230208073445.2317192-1-yoshihiro.shimoda.uh@renesas.com/ ==================== Link: https://lore.kernel.org/r/20230209081741.2536034-1-yoshihiro.shimoda.uh@renesas.comSigned-off-by:
-
Yoshihiro Shimoda authored
In the previous code, TX timestamp accuracy was bad because the irq handler got the timestamp from the timestamp register at that time. This hardware has "Timestamp capture" feature which can store each TX timestamp into the timestamp descriptors. To improve TX timestamp accuracy, implement timestamp descriptors' handling. Signed-off-by:
Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com> Reviewed-by:
Alexander Duyck <alexanderduyck@fb.com> Signed-off-by:
-
Yoshihiro Shimoda authored
In the previous code, the gptp flag was completely related to the !dir_tx in struct rswitch_gwca_queue because rswitch_gwca_queue_alloc() was called below: < In rswitch_txdmac_alloc() > err = rswitch_gwca_queue_alloc(ndev, priv, rdev->tx_queue, true, false, TX_RING_SIZE); So, dir_tx = true, and gptp = false. < In rswitch_rxdmac_alloc() > err = rswitch_gwca_queue_alloc(ndev, priv, rdev->rx_queue, false, true, RX_RING_SIZE); So, dir_tx = false, and gptp = true. In the future, a new queue handling for timestamp will be implemented and this gptp flag is confusable. So, remove the gptp flag. Signed-off-by:
-
Yoshihiro Shimoda authored
To improve readability, move linkfix related variables to struct rswitch_gwca. Also, rename function names "desc" with "linkfix". Signed-off-by:
-
Yoshihiro Shimoda authored
To add a new ring which is really related to timestamp (ts_ring) in the future, rename the following members to improve readability: ring --> tx_ring ts_ring --> rx_ring Signed-off-by: -
Dan Carpenter authored
This function always returns success. We need to preserve the error code before setting rx_ring->page_pool = NULL. Fixes: 850b9711 ("net: libwx: Allocate Rx and Tx resources") Signed-off-by:
Dan Carpenter <error27@gmail.com> Link: https://lore.kernel.org/r/Y+T4aoefc1XWvGYb@kiliSigned-off-by:
-
Arnd Bergmann authored
A new user of MSCC_OCELOT_SWITCH_LIB was added, bringing back an old link failure that was fixed with e5f31552 ("ethernet: fix PTP_1588_CLOCK dependencies"): x86_64-linux-ld: drivers/net/ethernet/mscc/ocelot_ptp.o: in function `ocelot_ptp_enable': ocelot_ptp.c:(.text+0x8ee): undefined reference to `ptp_find_pin' x86_64-linux-ld: drivers/net/ethernet/mscc/ocelot_ptp.o: in function `ocelot_get_ts_info': ocelot_ptp.c:(.text+0xd5d): undefined reference to `ptp_clock_index' x86_64-linux-ld: drivers/net/ethernet/mscc/ocelot_ptp.o: in function `ocelot_init_timestamp': ocelot_ptp.c:(.text+0x15ca): undefined reference to `ptp_clock_register' x86_64-linux-ld: drivers/net/ethernet/mscc/ocelot_ptp.o: in function `ocelot_deinit_timestamp': ocelot_ptp.c:(.text+0x16b7): undefined reference to `ptp_clock_unregister' Add the same PTP dependency here, as well as in the MSCC_OCELOT_SWITCH_LIB symbol itself to make it more obvious what is going on when the next driver selects it. Fixes: 3d7316ac ("net: dsa: ocelot: add external ocelot switch control") Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Acked-by:
Colin Foster <colin.foster@in-advantage.com> Link: https://lore.kernel.org/r/20230209124435.1317781-1-arnd@kernel.orgSigned-off-by:
-
Jakub Kicinski authored
Ido Schimmel says: ==================== bridge: mcast: Preparations for VXLAN MDB This patchset contains small preparations for VXLAN MDB that were split from this RFC [1]. Tested using existing bridge MDB forwarding selftests. [1] https://lore.kernel.org/netdev/20230204170801.3897900-1-idosch@nvidia.com/ ==================== Link: https://lore.kernel.org/r/20230209071852.613102-1-idosch@nvidia.comSigned-off-by:
-
Ido Schimmel authored
The kernel maintains three markers for the MDB dump: 1. The last bridge device from which the MDB was dumped. 2. The last MDB entry from which the MDB was dumped. 3. The last port-group entry that was dumped. Add test cases for large scale MDB dump to make sure that all the configured entries are dumped and that the markers are used correctly. Specifically, create 2 bridges with 32 ports and add 256 MDB entries in which all the ports are member of. Test that each bridge reports 8192 (256 * 32) permanent entries. Do that with IPv4, IPv6 and L2 MDB entries. On my system, MDB dump of the above is contained in about 50 netlink messages. Example output: # ./bridge_mdb.sh [...] INFO: # Large scale dump tests TEST: IPv4 large scale dump tests [ OK ] TEST: IPv6 large scale dump tests [ OK ] TEST: L2 large scale dump tests [ OK ] [...] Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Reviewed-by:
Petr Machata <petrm@nvidia.com> Acked-by:
Nikolay Aleksandrov <razor@blackwall.org> Signed-off-by:
-
Ido Schimmel authored
Future patches are going to move parts of the bridge MDB code to the common rtnetlink code in preparation for VXLAN MDB support. To facilitate code sharing between both drivers, move the validation of the top level attributes in RTM_{NEW,DEL}MDB messages to a policy that will eventually be moved to the rtnetlink code. Use 'NLA_NESTED' for 'MDBA_SET_ENTRY_ATTRS' instead of NLA_POLICY_NESTED() as this attribute is going to be validated using different policies in the underlying drivers. Signed-off-by:
-
