| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The tsnep network controller is able to extend the descriptor directly
with data to be transmitted. In this case no TX data DMA address is
necessary. Instead of the TX data DMA address the TX data buffer is
placed at the end of the descriptor.
The descriptor is read with a 64 bytes DMA read by the tsnep network
controller. If the sum of descriptor data and TX data is less than or
equal to 64 bytes, then no additional DMA read is necessary to read the
TX data. Therefore, it makes sense to inline small fragments up to this
limit within the descriptor ring.
Inlined fragments need to be copied to the descriptor ring. On the other
hand DMA mapping is not necessary. At most 40 bytes are copied, so
copying should be faster than DMA mapping.
For A53 1.2 GHz copying takes <100ns and DMA mapping takes >200ns. So
inlining small fragments should result in lower CPU load. Performance
improvement is small. Thus, comparision of CPU load with and without
inlining of small fragments did not show any significant difference.
With this optimization less DMA reads will be done, which decreases the
load of the interconnect.
Signed-off-by: Gerhard Engleder <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
The route lookup can be done now via generic function
udp_tunnel_dst_lookup() to replace the custom implementations in
vxlan_get_route().
Note that this patch only touches IPv4, while IPv6 still uses
vxlan6_get_route(). After IPv6 route lookup gets converted as well,
vxlan_xmit_one() can be simplified by removing local variables that
will be passed via "struct ip_tunnel_key", such as remote_ip,
local_ip, flow_flags, label.
Suggested-by: Guillaume Nault <[email protected]>
Signed-off-by: Beniamino Galvani <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
The route lookup can be done now via generic function
udp_tunnel_dst_lookup() to replace the custom implementation in
geneve_get_v4_rt().
Suggested-by: Guillaume Nault <[email protected]>
Signed-off-by: Beniamino Galvani <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Add a helper function to compute the tos/dsfield. In this way, we can
factor out some duplicate code. Also, the helper will be called from
more places in the next commit.
Suggested-by: Guillaume Nault <[email protected]>
Signed-off-by: Beniamino Galvani <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
We want to make the function more generic so that it can be used by
other UDP tunnel implementations such as geneve and vxlan. To do that,
add the following arguments:
- source and destination UDP port;
- ifindex of the output interface, needed by vxlan;
- the tos, because in some cases it is not taken from struct
ip_tunnel_info (for example, when it's inherited from the inner
packet);
- the dst cache, because not all tunnel types (e.g. vxlan) want to
use the one from struct ip_tunnel_info.
With these parameters, the function no longer needs the full struct
ip_tunnel_info as argument and we can pass only the relevant part of
it (struct ip_tunnel_key).
Suggested-by: Guillaume Nault <[email protected]>
Signed-off-by: Beniamino Galvani <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
The function is now UDP-specific, the protocol is always IPPROTO_UDP.
Suggested-by: Guillaume Nault <[email protected]>
Signed-off-by: Beniamino Galvani <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
At the moment ip_route_output_tunnel() is used only by bareudp.
Ideally, other UDP tunnel implementations should use it, but to do so
the function needs to accept new parameters that are specific for UDP
tunnels, such as the ports.
Prepare for these changes by renaming the function to
udp_tunnel_dst_lookup() and move it to file
net/ipv4/udp_tunnel_core.c.
Suggested-by: Guillaume Nault <[email protected]>
Signed-off-by: Beniamino Galvani <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Simplify logic for rspq_check_napi.
Drop redundant and wrong napi_is_scheduled call as it's not race free
and directly use the output of napi_schedule to understand if a napi is
pending or not.
rspq_check_napi main logic is to check if is_new_response is true and
check if a napi is not scheduled. The result of this function is then
used to detect if we are missing some interrupt and act on top of
this... With this knowing, we can rework and simplify the logic and make
it less problematic with testing an internal bit for napi.
Suggested-by: Eric Dumazet <[email protected]>
Signed-off-by: Christian Marangi <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Commit 295525e29a5b ("virtio_net: merge dma operations when filling
mergeable buffers") unmaps the buffer with DMA_ATTR_SKIP_CPU_SYNC when
the dma->ref is zero. We do that with DMA_ATTR_SKIP_CPU_SYNC, because we
do not want to do the sync for the entire page_frag. But that misses the
sync for the current area.
This patch does cpu sync regardless of whether the ref is zero or not.
Fixes: 295525e29a5b ("virtio_net: merge dma operations when filling mergeable buffers")
Reported-by: Michael Roth <[email protected]>
Closes: http://lore.kernel.org/all/[email protected]
Signed-off-by: Xuan Zhuo <[email protected]>
Acked-by: Jason Wang <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Implement new callback ops related to measurement and adjustment of
signal phase for pin-dpll in ice driver.
Signed-off-by: Arkadiusz Kubalewski <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Expose stored PBA ID string as unique board identifier via
devlink's .info_get command.
Signed-off-by: Ivan Vecera <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Function i40e_read_pba_string() is currently unused but will be used
by subsequent patch to provide board ID via devlink device info.
The function reads PBA block from NVM so it cannot be called during
adapter reset and as we would like to provide PBA ID via devlink
info it is better to read the PBA ID during i40e_probe() and cache
it in i40e_hw structure to avoid a waiting for potential adapter
reset in devlink info callback.
So...
- Remove pba_num and pba_num_size arguments from the function,
allocate resource managed buffer to store PBA ID string and
save resulting pointer to i40e_hw->pba_id field
- Make the function void as the PBA ID can be missing and in this
case (or in case of NVM reading failure) the i40e_hw->pba_id
will be NULL
- Rename the function to i40e_get_pba_string() to align with other
functions like i40e_get_oem_version() i40e_get_port_mac_addr()...
- Call this function on init during i40e_probe()
Signed-off-by: Ivan Vecera <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Provide devlink .info_get callback to allow the driver to report
detailed version information. The following info is reported:
"serial_number" -> The PCI DSN of the adapter
"fw.mgmt" -> The version of the firmware
"fw.mgmt.api" -> The API version of interface exposed over the AdminQ
"fw.psid" -> The version of the NVM image
"fw.bundle_id" -> Unique identifier for the combined flash image
"fw.undi" -> The combo image version
With this, 'devlink dev info' provides at least the same amount
information as is reported by ETHTOOL_GDRVINFO:
$ ethtool -i enp2s0f0 | egrep '(driver|firmware)'
driver: i40e
firmware-version: 9.30 0x8000e5f3 1.3429.0
$ devlink dev info pci/0000:02:00.0
pci/0000:02:00.0:
driver i40e
serial_number c0-de-b7-ff-ff-ef-ec-3c
versions:
running:
fw.mgmt 9.130.73618
fw.mgmt.api 1.15
fw.psid 9.30
fw.bundle_id 0x8000e5f3
fw.undi 1.3429.0
Signed-off-by: Ivan Vecera <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
The function formats NVM version string according adapter's
EETrackID value. If this value OEM specific (0xffffffff) then
the reported version is with format:
"<gen>.<snap>.<release>"
and in other case
"<nvm_maj>.<nvm_min> <eetrackid> <cvid_maj>.<cvid_bld>.<cvid_min>"
These versions are reported in the subsequent patch in this series
that implements devlink .info_get but separately.
So split the function into separate ones, refactor it to use them
and remove ugly static string buffer.
Additionally convert NVM/OEM version mask macros to use GENMASK and
use FIELD_GET/FIELD_PREP for them in i40e_nvm_version_str() and
i40e_get_oem_version(). This makes code more readable and allows
us to remove related shift macros.
Signed-off-by: Ivan Vecera <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Add an initial support for devlink interface to i40e driver.
Similarly to ice driver the implementation doe not enable devlink
to manage device-wide configuration and devlink instance is created
for each physical function of PCIe device.
Signed-off-by: Ivan Vecera <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
When we are trying to timestamp a TX packet, there may be
occasions when the TX timestamp register is still not
updated with the latest timestamp even if the timestamp
packet descriptor is marked as complete.
This usually happens in cases where the system is under
stress or flow control is affecting the transmit side.
We will solve this problem by saving the snapshot of the
timestamp register when we are posting the TX descriptor.
At this time, the register contains previously timestamped
packet's value and valid timestamp of the current packet must
be different than this.
Upon completion of the current descriptor, we will check if
the timestamp register is updated or not before timestamping
the skb. If not updated, we will schedule the ptp worker to
fetch the updated time later and timestamp the skb.
Also now we restrict number of outstanding PTP TX packet
requests to 1.
Reported-by: Simon White <[email protected]>
Link: https://lore.kernel.org/netdev/CACKFLikGdN9XPtWk-fdrzxdcD=+bv-GHBvfVfSpJzHY7hrW39g@mail.gmail.com/
Signed-off-by: Pavan Chebbi <[email protected]>
Signed-off-by: Michael Chan <[email protected]>
Reviewed-by: Andy Gospodarek <[email protected]>
Reviewed-by: Kalesh AP <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Driver allocates the LL2 rx buffers from kmalloc()
area to construct the skb using slab_build_skb()
The required size allocation seems to have overlooked
for accounting both skb_shared_info size and device
placement padding bytes which results into the below
panic when doing skb_put() for a standard MTU sized frame.
skbuff: skb_over_panic: text:ffffffffc0b0225f len:1514 put:1514
head:ff3dabceaf39c000 data:ff3dabceaf39c042 tail:0x62c end:0x566
dev:<NULL>
…
skb_panic+0x48/0x4a
skb_put.cold+0x10/0x10
qed_ll2b_complete_rx_packet+0x14f/0x260 [qed]
qed_ll2_rxq_handle_completion.constprop.0+0x169/0x200 [qed]
qed_ll2_rxq_completion+0xba/0x320 [qed]
qed_int_sp_dpc+0x1a7/0x1e0 [qed]
This patch fixes this by accouting skb_shared_info and device
placement padding size bytes when allocating the buffers.
Cc: David S. Miller <[email protected]>
Fixes: 0a7fb11c23c0 ("qed: Add Light L2 support")
Signed-off-by: Manish Chopra <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
If an IP address and/or L4 port for NAPT is available from a CT match,
the MAE will perform the edits; if no CT lookup has been performed for
this packet, the CT lookup did not return a match, or the matched CT
entry did not include NAPT, the action will have no effect.
Reviewed-by: Pieter Jansen van Vuuren <[email protected]>
Signed-off-by: Edward Cree <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
The MAE can edit either address, L4 port, or both, for either source
or destination. These can't be mixed; i.e. it can edit source addr
and source port, but not (say) source addr and dest port.
Reviewed-by: Pieter Jansen van Vuuren <[email protected]>
Signed-off-by: Edward Cree <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Actually the mlx5 code already has needed support to allow users
to configure soft/hard limits in bytes. It is possible due to the
situation with TX path, where CX7 devices are missing hardware
implementation to send events to the software, see commit b2f7b01d36a9
("net/mlx5e: Simulate missing IPsec TX limits hardware functionality").
That software workaround is not limited to TX and works for bytes too.
So relax the validation logic to not block soft/hard limits in bytes.
Reviewed-by: Patrisious Haddad <[email protected]>
Signed-off-by: Leon Romanovsky <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Increase max supported channels number to 256 (it is not extended
further due to testing disabilities).
Signed-off-by: Adham Faris <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Data center server CPUs number keeps getting larger with time.
Currently, our driver limits the number of channels to 128.
Maximum channels number is enforced and bounded by hardcoded
defines (en.h/MLX5E_MAX_NUM_CHANNELS) even though the device and machine
(CPUs num) can allow more.
Refactor current implementation in order to handle further channels.
The maximum supported channels number will be increased in the followup
patch.
Introduce RQT size calculation/allocation scheme below:
1) Preserve current RQT size of 256 for channels number up to 128 (the
old limit).
2) For greater channels number, RQT size is calculated by multiplying
the channels number by 2 and rounding up the result to the nearest
power of 2. If the calculated RQT size exceeds the maximum supported
size by the NIC, fallback to this maximum RQT size
(1 << log_max_rqt_size).
Since RQT size is no more static, allocate and free the indirection
table SW shadow dynamically.
Signed-off-by: Adham Faris <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Introduce code refactoring below:
1) Introduce single API for creating and destroying rss object,
mlx5e_rss_create() and mlx5e_rss_destroy() respectively.
2) mlx5e_rss_create() constructs and initializes RSS object depends
on a function new param enum mlx5e_rss_create_type. Callers (like
rx_res.c) will no longer need to allocate RSS object via
mlx5e_rss_alloc() and initialize it immediately via
mlx5e_rss_init_no_tirs() or mlx5e_rss_init(), this will be done by
a single call to mlx5e_rss_create(). Hence, mlx5e_rss_alloc() and
mlx5e_rss_init_no_tirs() have been removed from rss.h file and became
static functions.
Signed-off-by: Adham Faris <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Initialize indirect table array with memcpy rather than for loop.
This change has made for two reasons:
1) To be consistent with the indirect table array init in
mlx5e_rss_set_rxfh().
2) In general, prefer to use memcpy for array initializing rather than
for loop.
Signed-off-by: Adham Faris <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Refactor mlx5e_rx_res_init() and mlx5e_rx_res_free() by wrapping
mlx5e_rx_res_alloc() and mlx5e_rx_res_destroy() API's respectively.
Signed-off-by: Adham Faris <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Use the standard error pointer macro to shorten the code and simplify.
Signed-off-by: Yu Liao <[email protected]>
Reviewed-by: Leon Romanovsky <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Return PTR_ERR_OR_ZERO() instead of return 0 or PTR_ERR() to
simplify code.
Signed-off-by: Jinjie Ruan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Commit 2ac9cfe78223 ("net/mlx5e: IPSec, Add Innova IPSec offload TX data path")
declared mlx5e_ipsec_inverse_table_init() but never implemented it.
Commit f52f2faee581 ("net/mlx5e: Introduce flow steering API")
declared mlx5e_fs_set_tc() but never implemented it.
Commit f2f3df550139 ("net/mlx5: EQ, Privatize eq_table and friends")
declared mlx5_eq_comp_cpumask() but never implemented it.
Commit cac1eb2cf2e3 ("net/mlx5: Lag, properly lock eswitch if needed")
removed mlx5_lag_update() but not its declaration.
Commit 35ba005d820b ("net/mlx5: DR, Set flex parser for TNL_MPLS dynamically")
removed mlx5dr_ste_build_tnl_mpls() but not its declaration.
Commit e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
declared but never implemented mlx5_alloc_cmd_mailbox_chain() and mlx5_free_cmd_mailbox_chain().
Commit 0cf53c124756 ("net/mlx5: FWPage, Use async events chain")
removed mlx5_core_req_pages_handler() but not its declaration.
Commit 938fe83c8dcb ("net/mlx5_core: New device capabilities handling")
removed mlx5_query_odp_caps() but not its declaration.
Commit f6a8a19bb11b ("RDMA/netdev: Hoist alloc_netdev_mqs out of the driver")
removed mlx5_rdma_netdev_alloc() but not its declaration.
Signed-off-by: Yue Haibing <[email protected]>
Reviewed-by: Leon Romanovsky <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
mlx5_intf_lock is used to sync between LAG changes and its slaves
mlx5 core dev aux devices changes, which means every time mlx5 core
dev add/remove aux devices, mlx5 is taking this global lock, even if
LAG functionality isn't supported over the core dev.
This cause a bottleneck when probing VFs/SFs in parallel.
Hence, replace mlx5_intf_lock with HCA devcom component lock, or no
lock if LAG functionality isn't supported.
Signed-off-by: Shay Drory <[email protected]>
Reviewed-by: Mark Bloch <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
LAG peer device lookout bus logic required the usage of global lock,
mlx5_intf_mutex.
As part of the effort to remove this global lock, refactor LAG peer
device lookout to use mlx5 devcom layer.
Signed-off-by: Shay Drory <[email protected]>
Reviewed-by: Mark Bloch <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Downstream patch will add devcom component which will be locked in
many places. This can lead to a false positive "possible circular
locking dependency" warning by lockdep, on flows which lock more than
one mlx5 devcom component, such as probing ETH aux device.
Hence, add a lock_class_key per mlx5 device.
Signed-off-by: Shay Drory <[email protected]>
Reviewed-by: Mark Bloch <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
active_work is a work that iterates over all
possible SF devices which their SF port
representors are located on different function,
and in case SF is in active state, probes it.
Currently, the active_work in active_wq is
synced with mlx5_vhca_events_work via table_lock
and this lock causing a bottleneck in performance.
To remove table_lock, redesign active_wq logic
so that it now pushes active_work per SF to
mlx5_vhca_events_workqueues. Since the latter
workqueues are ordered, active_work and
mlx5_vhca_events_work with same index will be
pushed into same workqueue, thus it completely
eliminates the need for a lock.
Signed-off-by: Wei Zhang <[email protected]>
Signed-off-by: Shay Drory <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
At present, mlx5 driver have a general purpose
event handler which not only handles vhca event
but also many other events. This incurs a huge
bottleneck because the event handler is
implemented by single threaded workqueue and all
events are forced to be handled in serial manner
even though application tries to create multiple
SFs simultaneously.
Introduce a dedicated vhca event handler which
manages SFs parallel creation.
Signed-off-by: Wei Zhang <[email protected]>
Reviewed-by: Moshe Shemesh <[email protected]>
Reviewed-by: Shay Drory <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
|
|
Beamforming monitor is used to adjust registers to fine tune performance
and power save, and currently only existing WiFi 6 chips need it.
Signed-off-by: Zong-Zhe Yang <[email protected]>
Signed-off-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
When associated peer has beamformer capability, we should enable
beamformee, set CSI parameter, and configure rate to send CSI packets.
Since registers of WiFi 7 chips are very different from existing chips,
separate configuration functions.
Signed-off-by: Zong-Zhe Yang <[email protected]>
Signed-off-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
When associated peer has beamformer capability, enable hardware beamformee
function, and then hardware can run sounding protocol itself. Oppositely,
disable this function when disassociated. Define different registers for
WiFi 6 and 7 generations respectively.
Signed-off-by: Zong-Zhe Yang <[email protected]>
Signed-off-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
According to chip generation, set MU-EDCA parameters from mac80211 when
connected.
Signed-off-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
When connected with 802.11ax AP, MU-EDCA parameters are given, so enable
this hardware function by registers according to chip generation.
Signed-off-by: Zong-Zhe Yang <[email protected]>
Signed-off-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
When TX size or time of packet over RTS threshold set by this register,
hardware will use RTS protection automatically. Since WiFi 6 and 7 chips
have different register address for this, separate the address according
to chip gen.
Signed-off-by: Zong-Zhe Yang <[email protected]>
Signed-off-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
Since 'rtlpriv->cfg->ops->fill_tx_cmddesc()' is always called
with 'firstseg' and 'lastseg' set to 1 (and the latter is
never actually used), all of the relevant chip-specific
routines may be simplified. Compile tested only.
Signed-off-by: Dmitry Antipov <[email protected]>
Acked-by: Ping-Ke Shih <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
The ioctl handler has no actual callers in the kernel and is useless.
All the functionality should be reachable through the regualar interfaces.
Acked-by: Kalle Valo <[email protected]>
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
This function has no callers, and for the past 20 years, the request_firmware
interface has been in place instead of the custom firmware loader.
Acked-by: Kalle Valo <[email protected]>
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Kalle Valo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5 fixes 2023-10-12
This series provides bug fixes to mlx5 driver.
* tag 'mlx5-fixes-2023-10-12' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
net/mlx5e: Fix VF representors reporting zero counters to "ip -s" command
net/mlx5e: Don't offload internal port if filter device is out device
net/mlx5e: Take RTNL lock before triggering netdev notifiers
net/mlx5e: XDP, Fix XDP_REDIRECT mpwqe page fragment leaks on shutdown
net/mlx5e: RX, Fix page_pool allocation failure recovery for legacy rq
net/mlx5e: RX, Fix page_pool allocation failure recovery for striding rq
net/mlx5: Handle fw tracer change ownership event based on MTRC
net/mlx5: Bridge, fix peer entry ageing in LAG mode
net/mlx5: E-switch, register event handler before arming the event
net/mlx5: Perform DMA operations in the right locations
====================
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
One thing is broken in the safe mode, that is
ice_deinit_features() is being executed even
that ice_init_features() was not causing stack
trace during pci_unregister_driver().
Add check on the top of the function.
Fixes: 5b246e533d01 ("ice: split probe into smaller functions")
Signed-off-by: Mateusz Pacuszka <[email protected]>
Signed-off-by: Jan Sokolowski <[email protected]>
Reviewed-by: Przemek Kitszel <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
When the system boots into the crash dump kernel after a panic, the ice
networking device may still have pending transactions that can cause errors
or machine checks when the device is re-enabled. This can prevent the crash
dump kernel from loading the driver or collecting the crash data.
To avoid this issue, perform a function level reset (FLR) on the ice device
via PCIe config space before enabling it on the crash kernel. This will
clear any outstanding transactions and stop all queues and interrupts.
Restore the config space after the FLR, otherwise it was found in testing
that the driver wouldn't load successfully.
The following sequence causes the original issue:
- Load the ice driver with modprobe ice
- Enable SR-IOV with 2 VFs: echo 2 > /sys/class/net/eth0/device/sriov_num_vfs
- Trigger a crash with echo c > /proc/sysrq-trigger
- Load the ice driver again (or let it load automatically) with modprobe ice
- The system crashes again during pcim_enable_device()
Fixes: 837f08fdecbe ("ice: Add basic driver framework for Intel(R) E800 Series")
Reported-by: Vishal Agrawal <[email protected]>
Reviewed-by: Jay Vosburgh <[email protected]>
Reviewed-by: Przemek Kitszel <[email protected]>
Signed-off-by: Jesse Brandeburg <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
The hardware provides the indexes of the first and the last available
queue and VF. From the indexes, the driver calculates the numbers of
queues and VFs. In theory, a faulty device might say the last index is
smaller than the first index. In that case, the driver's calculation
would underflow, it would attempt to write to non-existent registers
outside of the ioremapped range and crash.
I ran into this not by having a faulty device, but by an operator error.
I accidentally ran a QE test meant for i40e devices on an ice device.
The test used 'echo i40e > /sys/...ice PCI device.../driver_override',
bound the driver to the device and crashed in one of the wr32 calls in
i40e_clear_hw.
Add checks to prevent underflows in the calculations of num_queues and
num_vfs. With this fix, the wrong device probing reports errors and
returns a failure without crashing.
Fixes: 838d41d92a90 ("i40e: clear all queues and interrupts")
Signed-off-by: Michal Schmidt <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
strncpy() is deprecated for use on NUL-terminated destination strings
[1] and as such we should prefer more robust and less ambiguous string
interfaces.
This patch eliminates three uses of strncpy():
Firstly, `dest` is expected to be NUL-terminated which is evident by the
manual setting of a NUL-byte at size - 1. For this use specifically,
strscpy() is a viable replacement due to the fact that it guarantees
NUL-termination on the destination buffer.
The next two cases should simply be memcpy() as the size of the src
string is always 3 and the destination string just wants the first 3
bytes changed.
To be clear, there are no buffer overread bugs in the current code as
the sizes and offsets are carefully managed such that buffers are
NUL-terminated. However, with these changes, the code is now more robust
and less ambiguous (and hopefully easier to read).
Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2]
Link: https://github.com/KSPP/linux/issues/90
Signed-off-by: Justin Stitt <[email protected]>
Reviewed-by: Kees Cook <[email protected]>
Link: https://lore.kernel.org/r/20231012-strncpy-drivers-net-ethernet-qlogic-qed-qed_debug-c-v2-1-16d2c0162b80@google.com
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
In very rare cases (I've seen two reports so far about different
RTL8125 chip versions) it seems the MAC locks up when link goes down
and requires a software reset to get revived.
Realtek doesn't publish hw errata information, therefore the root cause
is unknown. Realtek vendor drivers do a full hw re-initialization on
each link-up event, the slimmed-down variant here was reported to fix
the issue for the reporting user.
It's not fully clear which parts of the NIC are reset as part of the
software reset, therefore I can't rule out side effects.
Fixes: f1bce4ad2f1c ("r8169: add support for RTL8125")
Reported-by: Martin Kjær Jørgensen <[email protected]>
Link: https://lore.kernel.org/netdev/[email protected]/T/
Signed-off-by: Heiner Kallweit <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
ICSSG HW stats on TX side considers 8 preamble bytes as data bytes. Due
to this the tx_bytes of ICSSG interface doesn't match the rx_bytes of the
link partner. There is no public errata available yet.
As a workaround to fix this, decrease tx_bytes by 8 bytes for every tx
frame.
Fixes: c1e10d5dc7a1 ("net: ti: icssg-prueth: Add ICSSG Stats")
Signed-off-by: MD Danish Anwar <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Enable the attachment of a dynamic target to the target created during
boot time. The boot-time targets are named as "cmdline\d", where "\d" is
a number starting at 0.
If the user creates a dynamic target named "cmdline0", it will attach to
the first target created at boot time (as defined in the
`netconsole=...` command line argument). `cmdline1` will attach to the
second target and so forth.
If there is no netconsole target created at boot time, then, the target
name could be reused.
Relevant design discussion:
https://lore.kernel.org/all/[email protected]/
Suggested-by: Joel Becker <[email protected]>
Signed-off-by: Breno Leitao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|