| Age | Commit message (Collapse) | Author |
|---|
|
Before this patch the NETDEV_XDP_ACT_NDO_XMIT XDP feature flag is set by
default as part of driver initialization, and is never cleared. However,
this flag differs from others in that it is used as an indicator for
whether the driver is ready to perform the ndo_xdp_xmit operation as
part of an XDP_REDIRECT. Kernel helpers
xdp_features_(set|clear)_redirect_target exist to convey this meaning.
This patch ensures that the netdev is only reported as a redirect target
when XDP queues exist to forward traffic.
Fixes: 39a7f4aa3e4a ("gve: Add XDP REDIRECT support for GQI-QPL format")
Cc: stable@vger.kernel.org
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Link: https://patch.msgid.link/20250214224417.1237818-1-joshwash@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Commit f85949f98206 ("xdp: add xdp_set_features_flag utility routine")
added routines to inform the core about XDP flag changes.
GVE support was added around the same time and missed using them.
GVE only changes the flags on error recover or resume.
Presumably the flags may change during resume if VM migrated.
User would not get the notification and upper devices would
not get a chance to recalculate their flags.
Fixes: 75eaae158b1b ("gve: Add XDP DROP and TX support for GQI-QPL format")
Reviewed-By: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20250106180210.1861784-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Commit ba0925c34e0f ("gve: process XSK TX descriptors as part of RX NAPI")
moved XSK TX processing to be part of the RX NAPI. However, that commit
did not include triggering the RX NAPI in gve_xsk_wakeup. This is
necessary because the TX NAPI only processes TX completions, meaning
that a TX wakeup would not actually trigger XSK descriptor processing.
Also, the branch on XDP_WAKEUP_TX was supposed to have been removed, as
the NAPI should be scheduled whether the wakeup is for RX or TX.
Fixes: ba0925c34e0f ("gve: process XSK TX descriptors as part of RX NAPI")
Cc: stable@vger.kernel.org
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Link: https://patch.msgid.link/20241221032807.302244-1-pkaligineedi@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This patch fixes a number of consistency issues in the queue allocation
path related to XDP.
As it stands, the number of allocated XDP queues changes in three
different scenarios.
1) Adding an XDP program while the interface is up via
gve_add_xdp_queues
2) Removing an XDP program while the interface is up via
gve_remove_xdp_queues
3) After queues have been allocated and the old queue memory has been
removed in gve_queues_start.
However, the requirement for the interface to be up for
gve_(add|remove)_xdp_queues to be called, in conjunction with the fact
that the number of queues stored in priv isn't updated until _after_ XDP
queues have been allocated in the normal queue allocation path means
that if an XDP program is added while the interface is down, XDP queues
won't be added until the _second_ if_up, not the first.
Given the expectation that the number of XDP queues is equal to the
number of RX queues, scenario (3) has another problematic implication.
When changing the number of queues while an XDP program is loaded, the
number of XDP queues must be updated as well, as there is logic in the
driver (gve_xdp_tx_queue_id()) which relies on every RX queue having a
corresponding XDP TX queue. However, the number of XDP queues stored in
priv would not be updated until _after_ a close/open leading to a
mismatch in the number of XDP queues reported vs the number of XDP
queues which actually exist after the queue count update completes.
This patch remedies these issues by doing the following:
1) The allocation config getter function is set up to retrieve the
_expected_ number of XDP queues to allocate instead of relying
on the value stored in `priv` which is only updated once the queues
have been allocated.
2) When adjusting queues, XDP queues are adjusted to match the number of
RX queues when XDP is enabled. This only works in the case when
queues are live, so part (1) of the fix must still be available in
the case that queues are adjusted when there is an XDP program and
the interface is down.
Fixes: 5f08cd3d6423 ("gve: Alloc before freeing when adjusting queues")
Cc: stable@vger.kernel.org
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When busy polling is enabled, xsk_sendmsg for AF_XDP zero copy marks
the NAPI ID corresponding to the memory pool allocated for the socket.
In GVE, this NAPI ID will never correspond to a NAPI ID of one of the
dedicated XDP TX queues registered with the umem because XDP TX is not
set up to share a NAPI with a corresponding RX queue.
This patch moves XSK TX descriptor processing from the TX NAPI to the RX
NAPI, and the gve_xsk_wakeup callback is updated to use the RX NAPI
instead of the TX NAPI, accordingly. The branch on if the wakeup is for
TX is removed, as the NAPI poll should be invoked whether the wakeup is
for TX or for RX.
Fixes: fd8e40321a12 ("gve: Add AF_XDP zero-copy support for GQI-QPL format")
Cc: stable@vger.kernel.org
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch predicates the enabling and disabling of XSK pools on the
existence of queues. As it stands, if the interface is down, disabling
or enabling XSK pools would result in a crash, as the RX queue pointer
would be NULL. XSK pool registration will occur as part of the next
interface up.
Similarly, xsk_wakeup needs be guarded against queues disappearing
while the function is executing, so a check against the
GVE_PRIV_FLAGS_NAPI_ENABLED flag is added to synchronize with the
disabling of the bit and the synchronize_net() in gve_turndown.
Fixes: fd8e40321a12 ("gve: Add AF_XDP zero-copy support for GQI-QPL format")
Cc: stable@vger.kernel.org
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Larysa Zaremba <larysa.zaremba@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In GVE, dedicated XDP queues only exist when an XDP program is installed
and the interface is up. As such, the NDO XDP XMIT callback should
return early if either of these conditions are false.
In the case of no loaded XDP program, priv->num_xdp_queues=0 which can
cause a divide-by-zero error, and in the case of interface down,
num_xdp_queues remains untouched to persist XDP queue count for the next
interface up, but the TX pointer itself would be NULL.
The XDP xmit callback also needs to synchronize with a device
transitioning from open to close. This synchronization will happen via
the GVE_PRIV_FLAGS_NAPI_ENABLED bit along with a synchronize_net() call,
which waits for any RCU critical sections at call-time to complete.
Fixes: 39a7f4aa3e4a ("gve: Add XDP REDIRECT support for GQI-QPL format")
Cc: stable@vger.kernel.org
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implement netdev_stats_ops to export basic per-queue stats.
With page pool support for DQO added in the previous patches,
rx-alloc-fail captures failures in page pool allocations as
well since the rx_buf_alloc_fail stat tracked in the driver
is incremented when gve_alloc_buffer returns error.
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20241014202108.1051963-4-pkaligineedi@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use the netdev-genl interface to map NAPI instances to queues so that
this information is accessible to user programs via netlink.
$ ./tools/net/ynl/cli.py --spec Documentation/netlink/specs/netdev.yaml \
--dump queue-get --json='{"ifindex": 2}'
[{'id': 0, 'ifindex': 2, 'napi-id': 8313, 'type': 'rx'},
{'id': 1, 'ifindex': 2, 'napi-id': 8314, 'type': 'rx'},
{'id': 2, 'ifindex': 2, 'napi-id': 8315, 'type': 'rx'},
{'id': 3, 'ifindex': 2, 'napi-id': 8316, 'type': 'rx'},
{'id': 4, 'ifindex': 2, 'napi-id': 8317, 'type': 'rx'},
[...]
{'id': 0, 'ifindex': 2, 'napi-id': 8297, 'type': 'tx'},
{'id': 1, 'ifindex': 2, 'napi-id': 8298, 'type': 'tx'},
{'id': 2, 'ifindex': 2, 'napi-id': 8299, 'type': 'tx'},
{'id': 3, 'ifindex': 2, 'napi-id': 8300, 'type': 'tx'},
{'id': 4, 'ifindex': 2, 'napi-id': 8301, 'type': 'tx'},
[...]
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Link: https://patch.msgid.link/20240930210731.1629-3-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
GVE driver wrongly relies on netif_carrier_ok() to check the
interface administrative state when resources are being
allocated/deallocated for queue(s). netif_carrier_ok() needs
to be replaced with netif_running() for all such cases.
Administrative state is the result of "ip link set dev <dev>
up/down". It reflects whether the administrator wants to use
the device for traffic and the corresponding resources have
been allocated.
Fixes: 5f08cd3d6423 ("gve: Alloc before freeing when adjusting queues")
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20240801205619.987396-1-pkaligineedi@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Implement the ethtool commands that can be used to configure and query
flow-steering rules.
A large part of this change consists of translating the ethtool
representation of 'ntuples' to our internal gve_flow_rule and vice-versa
in the new created gve_flow_rule.c
Considering the possible large amount of flow rules, the driver doesn't
store all the rules locally. When the user runs 'ethtool -n <nic>' to
check the registered rules, the driver will send adminq command to
query a limited amount of rules/rule ids(that filled in a 4096 bytes dma
memory) at a time as a cache for the ethtool queries. The adminq query
commands will be repeated for several times until the ethtool has
queried all the needed rules.
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Co-developed-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20240625001232.1476315-6-ziweixiao@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add new adminq commands for the driver to configure and query flow rules
that are stored in the device. Flow steering rules are assigned with a
location that determines the relative order of the rules.
Flow rules can run up to an order of millions. In such cases, storing
a full copy of the rules in the driver to prepare for the ethtool query
is infeasible while querying them from the device is better. That needs
to be optimized too so that we don't send a lot of adminq commands. The
solution here is to store a limited number of rules/rule ids in the
driver in a cache. Use dma_pool to allocate 4k bytes which lets device
write at most 46 flow rules(4096/88) or 1024 rule ids(4096/4) at a time.
For configuring flow rules, there are 3 sub-commands:
- ADD which adds a rule at the location supplied
- DEL which deletes the rule at the location supplied
- RESET which clears all currently active rules in the device
For querying flow rules, there are also 3 sub-commands:
- QUERY_RULES corresponds to ETHTOOL_GRXCLSRULE. It fills the rules in
the allocated cache after querying the device
- QUERY_RULES_IDS corresponds to ETHTOOL_GRXCLSRLALL. It fills the
rule_ids in the allocated cache after querying the device
- QUERY_RULES_STATS corresponds to ETHTOOL_GRXCLSRLCNT. It queries the
device's current flow rule number and the supported max flow rule
limit
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Co-developed-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20240625001232.1476315-5-ziweixiao@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The new netdev queue api is implemented for gve.
Tested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Shailend Chand <shailend@google.com>
Link: https://lore.kernel.org/all/20240501232549.1327174-11-shailend@google.com/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Every tx and rx ring has its own queue-page-list (QPL) that serves as
the bounce buffer. Previously we were allocating QPLs for all queues
before the queues themselves were allocated and later associating a QPL
with a queue. This is avoidable complexity: it is much more natural for
each queue to allocate and free its own QPL.
Moreover, the advent of new queue-manipulating ndo hooks make it hard to
keep things as is: we would need to transfer a QPL from an old queue to
a new queue, and that is unpleasant.
Tested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In order to make possible the implementation of per-queue ndo hooks,
gve_turnup was changed in a previous patch to account for queues already
having some unprocessed descriptors: it does a one-off napi_schdule to
handle them. If conditions of consistent high traffic persist in the
immediate aftermath of this, the poll routine for a queue can be "stuck"
on the cpu on which the ndo hooks ran, instead of the cpu its irq has
affinity with.
This situation is exacerbated by the fact that the ndo hooks for all the
queues are invoked on the same cpu, potentially causing all the napi
poll routines to be residing on the same cpu.
A self correcting mechanism in the poll method itself solves this
problem.
Tested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
gVNIC has a requirement that all queues have to be quiesced before any
queue is operated on (created or destroyed). To enable the
implementation of future ndo hooks that work on a single queue, we need
to evolve gve_turnup to account for queues already having some
unprocessed descriptors in the ring.
Say rxq 4 is being stopped and started via the queue api. Due to gve's
requirement of quiescence, queues 0 through 3 are not processing their
rings while queue 4 is being toggled. Once they are made live, these
queues need to be poked to cause them to check their rings for
descriptors that were written during their brief period of quiescence.
Tested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently the queues are either all live or all dead, toggling from one
state to the other via the ndo open and stop hooks. The future addition
of single-queue ndo hooks changes this, and thus gve_turnup and
gve_turndown should evolve to account for a state where some queues are
live and some aren't.
Tested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The qpl_cfg struct was used to make sure that no two different queues
are using QPL with the same qpl_id. We can remove that qpl_cfg struct
since now the qpl_ids map with the queues respectively as follows:
For tx queues: qpl_id = tx_qid
For rx queues: qpl_id = max_tx_queues + rx_qid
And when XDP is used, it will need the user to reduce the tx queues to
be at most half of the max_tx_queues. Then it will use the same number
of tx queues starting from the end of existing tx queues for XDP. So the
XDP queues will not exceed the max_tx_queues range and will not overlap
with the rx queues, where the qpl_ids will not have overlapping too.
Considering of that, we remove the qpl_cfg struct to get the qpl_id
directly based on the queue id. Unless we are erroneously allocating a
rx/tx queue that has already been allocated, we would never allocate
the qpl with the same qpl_id twice. In that case, it should fail much
earlier than the QPL assignment.
Suggested-by: Praveen Kaligineedi <pkaligineedi@google.com>
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Shailend Chand <shailend@google.com>
Link: https://lore.kernel.org/r/20240417205757.778551-1-ziweixiao@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Allow the user to change ring size via ethtool if
supported by the device. The driver relies on the
ring size ranges queried from device to validate
ring sizes requested by the user.
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fulfill the requirement that for GQI, the number of pages per
RX QPL is equal to the ring size. Set this value to be equal to
ring size. Because of this change, the rx_data_slot_cnt and
rx_pages_per_qpl fields stored in the priv structure are not
needed, so remove their usage. And for DQO, the number of pages
per RX QPL is more than ring size to account for out-of-order
completions. So set it to two times of rx ring size.
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When draining a page_frag_cache, most user are doing
the similar steps, so introduce an API to avoid code
duplication.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Add header buffers and ethtool support to enable header split via the
tcp-data-split flag in ethtool's ringparam config. A coherent dma memory
is allocated for the header buffers. There is one header buffer per ring
entry by calculating the offset to the header-buffers starting address.
The header buffer is always copied directly into the skb and payload is
always added as frags. When there is a header buffer overflow or the
header length is 0, the driver places the whole unsplit packet in frags.
When toggling header split, the driver will call gve_adjust_config to
set its queues appropriately. If header split is enabled by the user and
the max packet buffer size is no less than 4KB, driver will set the
packet buffer size as 4KB to support TCP_ZEROCOPY_RECEIVE. Otherwise the
driver will use the default 2KB as the packet buffer size.
`ethtool -G <dev> tcp-data-split on/off` is the command to toggle header
split.
`ethtool -g <dev>` will show the status of header split with the field
of `tcp-data-split`.
Co-developed-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
To enable header split via ethtool, we first need to query the device to
get the max rx buffer size and header buffer size. Add a device option
to get these values and store them in the driver. If the header buffer
size received from the device is non-zero, it means header split is
supported in the device.
Currently the max rx buffer size will only be used when header split is
enabled which will set the data_buffer_size_dqo to be the max rx buffer
size. Also change the data_buffer_size_dqo from int to u16 since we are
modifying it and making it to be consistent with max_rx_buffer_size.
Co-developed-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Previously, existing queues were being freed before the resources for
the new queues were being allocated. This would take down the interface
if someone were to attempt to change feature flags under a resource
crunch.
Signed-off-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Link: https://lore.kernel.org/r/20240122182632.1102721-7-shailend@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Previously, existing queues were being freed before the resources for
the new queues were being allocated. This would take down the interface
if someone were to attempt to change queue counts under a resource
crunch.
Signed-off-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Link: https://lore.kernel.org/r/20240122182632.1102721-6-shailend@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
gve_open is rewritten to be composed of two funcs: gve_queues_mem_alloc
and gve_queues_start. The former only allocates queue resources without
doing anything to install the queues, which is taken up by the latter.
Similarly gve_close is split into gve_queues_stop and
gve_queues_mem_free.
Separating the acts of queue resource allocation and making the queue
become live help with subsequent changes that aim to not take down the
datapath when applying new configurations.
Signed-off-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Link: https://lore.kernel.org/r/20240122182632.1102721-5-shailend@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The new config-aware functions will help achieve the goal of being able
to allocate resources for new queues while there already are active
queues serving traffic.
These new functions work off of arbitrary queue allocation configs
rather than just the currently active config in priv, and they return
the newly allocated resources instead of writing them into priv.
Signed-off-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240122182632.1102721-4-shailend@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This change makes the napi poll functions non-static and moves the
gve_(add|remove)_napi functions to gve_utils.c, to make possible future
"start queue" hooks in the datapath files.
Signed-off-by: Shailend Chand <shailend@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240122182632.1102721-3-shailend@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Prior to this change, gve crashes when attempting to run in kernels with
page sizes other than 4k. This change removes unnecessary references to
PAGE_SIZE and replaces them with more meaningful constants.
Signed-off-by: Jordan Kimbrough <jrkim@google.com>
Signed-off-by: John Fraker <jfraker@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20231128002648.320892-6-jfraker@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
It is suboptimal to attempt skb linearization from ndo_start_xmit()
if a gso skb has pathological layout, or if host stack does not have
access to the payload (TCP direct). Linearization of large skbs
can also fail under memory pressure.
We should instead have an ndo_features_check() so that we can
fallback to GSO, which is supported even for TCP direct,
and generally much more efficient (no payload copy).
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Bailey Forrest <bcf@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Jeroen de Borst <jeroendb@google.com>
Cc: Praveen Kaligineedi <pkaligineedi@google.com>
Cc: Shailend Chand <shailend@google.com>
Cc: Ziwei Xiao <ziweixiao@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Netpoll will explicilty pass the polling call with a budget of 0 to
indicate it's clearing the Tx path only. For the gve_rx_poll and
gve_xdp_poll, they were mistakenly taking the 0 budget as the indication
to do all the work. Add check to avoid the rx path and xdp path being
called when budget is 0. And also avoid napi_complete_done being called
when budget is 0 for netpoll.
Fixes: f5cedc84a30d ("gve: Add transmit and receive support")
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Link: https://lore.kernel.org/r/20231114004144.2022268-1-ziweixiao@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Now that napi_schedule return a bool, we can drop napi_reschedule that
does the same exact function. The function comes from a very old commit
bfe13f54f502 ("ibm_emac: Convert to use napi_struct independent of struct
net_device") and the purpose is actually deprecated in favour of
different logic.
Convert every user of napi_reschedule to napi_schedule.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com> # ath10k
Acked-by: Nick Child <nnac123@linux.ibm.com> # ibm
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de> # for can/dev/rx-offload.c
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20231009133754.9834-3-ansuelsmth@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
If, for any reason, `tx_stats_num + rx_stats_num` wraps around, the
protection that struct_size() adds against potential integer overflows
is defeated. Fix this by hardening call to struct_size() with size_add().
Fixes: 691f4077d560 ("gve: Replace zero-length array with flexible-array member")
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
GVE supports QPL ("queue-page-list") mode where
all data is communicated through a set of pre-registered
pages. Adding this mode to DQO descriptor format.
Add checks, abi-changes and device options to support
QPL mode for DQO in addition to GQI. Also, use
pages-per-qpl supplied by device-option to control the
size of the "queue-page-list".
Signed-off-by: Rushil Gupta <rushilg@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Signed-off-by: Bailey Forrest <bcf@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Current codebase contained the usage of two different names for this
driver (i.e., `gvnic` and `gve`), which is quite unfriendly for users
to use, especially when trying to bind or unbind the driver manually.
The corresponding kernel module is registered with the name of `gve`.
It's more reasonable to align the name of the driver with the module.
Fixes: 893ce44df565 ("gve: Add basic driver framework for Compute Engine Virtual NIC")
Cc: csully@google.com
Signed-off-by: Junfeng Guo <junfeng.guo@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for using IPv6 Big TCP on DQ which can handle large TSO/GRO
packets. See https://lwn.net/Articles/895398/. This can improve the
throughput and CPU usage.
Perf test result:
ip -d link show $DEV
gso_max_size 185000 gso_max_segs 65535 tso_max_size 262143 tso_max_segs 65535 gro_max_size 185000
For performance, tested with neper using 9k MTU on hardware that supports 200Gb/s line rate.
In single streams when line rate is not saturated, we expect throughput improvements.
When the networking is performing at line rate, we expect cpu usage improvements.
Tcp_stream (unidirectional stream test, T=thread, F=flow):
skb=180kb, T=1, F=1, no zerocopy: throughput average=64576.88 Mb/s, sender stime=8.3, receiver stime=10.68
skb=64kb, T=1, F=1, no zerocopy: throughput average=64862.54 Mb/s, sender stime=9.96, receiver stime=12.67
skb=180kb, T=1, F=1, yes zerocopy: throughput average=146604.97 Mb/s, sender stime=10.61, receiver stime=5.52
skb=64kb, T=1, F=1, yes zerocopy: throughput average=131357.78 Mb/s, sender stime=12.11, receiver stime=12.25
skb=180kb, T=20, F=100, no zerocopy: throughput average=182411.37 Mb/s, sender stime=41.62, receiver stime=79.4
skb=64kb, T=20, F=100, no zerocopy: throughput average=182892.02 Mb/s, sender stime=57.39, receiver stime=72.69
skb=180kb, T=20, F=100, yes zerocopy: throughput average=182337.65 Mb/s, sender stime=27.94, receiver stime=39.7
skb=64kb, T=20, F=100, yes zerocopy: throughput average=182144.20 Mb/s, sender stime=47.06, receiver stime=39.01
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Coco Li <lixiaoyan@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230522201552.3585421-1-ziweixiao@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Clearing the PBA bit from the driver is race prone and it may lead to
dropped interrupt events. This could potentially lead to the traffic
being completely halted.
Fixes: 5e8c5adf95f8 ("gve: DQO: Add core netdev features")
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Signed-off-by: Bailey Forrest <bcf@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Adding AF_XDP zero-copy support.
Note: Although these changes support AF_XDP socket in zero-copy
mode, there is still a copy happening within the driver between
XSK buffer pool and QPL bounce buffers in GQI-QPL format.
In GQI-QPL queue format, the driver needs to allocate a fixed size
memory, the size specified by vNIC device, for RX/TX and register this
memory as a bounce buffer with the vNIC device when a queue is
created. The number of pages in the bounce buffer is limited and the
pages need to be made available to the vNIC by copying the RX data out
to prevent head-of-line blocking. Therefore, we cannot pass the XSK
buffer pool to the vNIC.
The number of copies on RX path from the bounce buffer to XSK buffer is 2
for AF_XDP copy mode (bounce buffer -> allocated page frag -> XSK buffer)
and 1 for AF_XDP zero-copy mode (bounce buffer -> XSK buffer).
This patch contains the following changes:
1) Enable and disable XSK buffer pool
2) Copy XDP packets from QPL bounce buffers to XSK buffer on rx
3) Copy XDP packets from XSK buffer to QPL bounce buffers and
ring the doorbell as part of XDP TX napi poll
4) ndo_xsk_wakeup callback support
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch contains the following changes:
1) Support for XDP REDIRECT action on rx
2) ndo_xdp_xmit callback support
In GQI-QPL queue format, the driver needs to allocate a fixed size
memory, the size specified by vNIC device, for RX/TX and register this
memory as a bounce buffer with the vNIC device when a queue is created.
The number of pages in the bounce buffer is limited and the pages need to
be made available to the vNIC by copying the RX data out to prevent
head-of-line blocking. The XDP_REDIRECT packets are therefore immediately
copied to a newly allocated page.
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for XDP PASS, DROP and TX actions.
This patch contains the following changes:
1) Support installing/uninstalling XDP program
2) Add dedicated XDP TX queues
3) Add support for XDP DROP action
4) Add support for XDP TX action
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Changes to enable adding and removing TX queues without calling
gve_close() and gve_open().
Made the following changes:
1) priv->tx, priv->rx and priv->qpls arrays are allocated based on
max tx queues and max rx queues
2) Changed gve_adminq_create_tx_queues(), gve_adminq_destroy_tx_queues(),
gve_tx_alloc_rings() and gve_tx_free_rings() functions to add/remove a
subset of TX queues rather than all the TX queues.
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds/modifies helper functions needed to add XDP
support.
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
IRQs are currently requested before the netdevice is registered
and a proper name is assigned to the device. Changing interrupt
name to avoid using the format string in the name.
Interrupt name before change: eth%d-ntfy-block.<blk_id>
Interrupt name after change: gve-ntfy-blk<blk_id>@pci:<pci_name>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Check whether the driver is compatible with the device
presented.
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Now that the 32bit UP oddity is gone and 32bit uses always a sequence
count, there is no need for the fetch_irq() variants anymore.
Convert to the regular interface.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
We tell driver developers to always pass NAPI_POLL_WEIGHT
as the weight to netif_napi_add(). This may be confusing
to newcomers, drop the weight argument, those who really
need to tweak the weight can use netif_napi_add_weight().
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de> # for CAN
Link: https://lore.kernel.org/r/20220927132753.750069-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
On 32bit-UP u64_stats_fetch_begin() disables only preemption. If the
reader is in preemptible context and the writer side
(u64_stats_update_begin*()) runs in an interrupt context (IRQ or
softirq) then the writer can update the stats during the read operation.
This update remains undetected.
Use u64_stats_fetch_begin_irq() to ensure the stats fetch on 32bit-UP
are not interrupted by a writer. 32bit-SMP remains unaffected by this
change.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Catherine Sullivan <csully@google.com>
Cc: David Awogbemila <awogbemila@google.com>
Cc: Dimitris Michailidis <dmichail@fungible.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Hans Ulli Kroll <ulli.kroll@googlemail.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Jeroen de Borst <jeroendb@google.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Simon Horman <simon.horman@corigine.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-wireless@vger.kernel.org
Cc: netdev@vger.kernel.org
Cc: oss-drivers@corigine.com
Cc: stable@vger.kernel.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The commit
a5886ef4f4bf ("gve: Introduce per netdev `enum gve_queue_format`")
introduces three queue format type, only GVE_GQI_QPL_FORMAT queue has
page list. So it should use the queue page list number to detect the
zero size queue page list. Correct the design logic.
Using the 'queue_format == GVE_GQI_RDA_FORMAT' may lead to request zero
sized memory allocation, like if the queue format is GVE_DQO_RDA_FORMAT.
The kernel memory subsystem will return ZERO_SIZE_PTR, which is not NULL
address, so the driver can run successfully. Also the code still checks
the queue page list number firstly, then accesses the allocated memory,
so zero number queue page list allocation will not lead to access fault.
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Reviewed-by: Bailey Forrest <bcf@google.com>
Link: https://lore.kernel.org/r/20220215051751.260866-1-haiyue.wang@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use GFP_ATOMIC when allocating pages out of the hotpath,
continue to use GFP_KERNEL when allocating pages during setup.
GFP_KERNEL will allow blocking which allows it to succeed
more often in a low memory enviornment but in the hotpath we do
not want to allow the allocation to block.
Fixes: f5cedc84a30d2 ("gve: Add transmit and receive support")
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Link: https://lore.kernel.org/r/20220126003843.3584521-1-awogbemila@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Adding ethtool support for changing rx-coalesce-usec and tx-coalesce-usec
when using the DQO queue format.
Signed-off-by: Tao Liu <xliutaox@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
