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>
This does not fix any existing bug. In anticipation of the ndo queue api
hooks that alloc/free/start/stop a single Rx queue, the already existing
per-queue stop functions are being made more robust. Specifically for
this use case: rx_queue_n.stop() + rx_queue_n.start()
Note that this is not the use case being used in devmem tcp (the first
place these new ndo hooks would be used). There the usecase is:
new_queue.alloc() + old_queue.stop() + new_queue.start() + old_queue.free()
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>
Although this is not fixing any existing double free bug, making these
functions idempotent allows for a simpler implementation of future ndo
hooks that act on a single queue.
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>
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>
Previously, each caller of gve_rx_alloc_buffer had to increase counter
and as a result one caller was not tracking those failure. Increasing
counters at a common location now so callers don't have to duplicate
code or miss counter management.
Signed-off-by: Ankit Garg <nktgrg@google.com>
Link: https://lore.kernel.org/r/20240124205435.1021490-1-nktgrg@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For a skb frag with a newly allocated copy page, the true size is
incorrectly set to packet buffer size. It should be set to PAGE_SIZE
instead.
Fixes: 82fd151d38 ("gve: Reduce alloc and copy costs in the GQ rx path")
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Link: https://lore.kernel.org/r/20240124161025.1819836-1-pkaligineedi@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>
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>
These checks are safe to remove as they are no longer enforced by the
backend. Retaining them would require updating them to work differently
with page sizes larger than 4k.
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/20231128002648.320892-4-jfraker@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
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: f5cedc84a3 ("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>
The prefill function should have only removed the page count bias it
added. Fully freeing the page will cause gve_free_queue_page_list to
free a page the driver no longer owns.
Fixes: 82fd151d38 ("gve: Reduce alloc and copy costs in the GQ rx path")
Signed-off-by: Shailend Chand <shailend@google.com>
Link: https://lore.kernel.org/r/20231014014121.2843922-1-shailend@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
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>
If alloc_page() fails in gve_prefill_rx_pages(), it should return
an error code in the error path.
Fixes: 82fd151d38 ("gve: Reduce alloc and copy costs in the GQ rx path")
Cc: Jeroen de Borst <jeroendb@google.com>
Cc: Catherine Sullivan <csully@google.com>
Cc: Shailend Chand <shailend@google.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously, even if just one of the many fragments of a 9k packet
required a copy, we'd copy the whole packet into a freshly-allocated
9k-sized linear SKB, and this led to performance issues.
By having a pool of pages to copy into, each fragment can be
independently handled, leading to a reduced incidence of
allocation and copy.
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This caused a significant performance degredation when using generic XDP
with multiple queues.
Fixes: f5cedc84a3 ("gve: Add transmit and receive support")
Signed-off-by: Tao Liu <xliutaox@google.com>
Link: https://lore.kernel.org/r/20220207175901.2486596-1-jeroendb@google.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: f5cedc84a3 ("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>
No longer needed after we introduced the barrier in gve_napi_poll.
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The u64_stats_update_end() call is supposed to be inside the curly
braces so it pairs with the u64_stats_update_begin().
Fixes: 37149e9374 ("gve: Implement packet continuation for RX.")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This enables the driver to receive RX packets spread across multiple
buffers:
For a given multi-fragment packet the "packet continuation" bit is set
on all descriptors except the last one. These descriptors' payloads are
combined into a single SKB before the SKB is handed to the
networking stack.
This change adds a "packet buffer size" notion for RX queues. The
CreateRxQueue AdminQueue command sent to the device now includes the
packet_buffer_size.
We opt for a packet_buffer_size of PAGE_SIZE / 2 to give the
driver the opportunity to flip pages where we can instead of copying.
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The rx_buf_alloc_fail counter wasn't getting updated.
Fixes: 433e274b8f ("gve: Add stats for gve.")
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Half pages are just used for small enough packets. This change allows
this to also apply for systems with pages larger than 4 KB.
Fixes: 02b0e0c18b ("gve: Rx Buffer Recycling")
Signed-off-by: Jordan Kim <jrkim@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a pagecnt bias field to rx buffer info struct to eliminate
needing to increment the atomic page ref count on every pass in the
rx hotpath.
Also prefetch two packet pages ahead.
Fixes: ede3fcf5ec ("gve: Add support for raw addressing to the rx path")
Signed-off-by: Yanchun Fu <yangchun@google.com>
Signed-off-by: Nathan Lewis <npl@google.com>
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use napi_complete_done to allow for the use of gro_flush_timeout.
Fixes: f5cedc84a3 ("gve: Add transmit and receive support")
Signed-off-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ignored errors would result in crash.
Fixes: ede3fcf5ec ("gve: Add support for raw addressing to the rx path")
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allocate the buffer and completion ring structures. Do not populate the
rings yet. That will happen in the respective rx and tx datapath
follow-on patches
Signed-off-by: Bailey Forrest <bcf@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The currently supported queue formats are:
- GQI_RDA - GQI with raw DMA addressing
- GQI_QPL - GQI with queue page list
- DQO_RDA - DQO with raw DMA addressing
The old `gve_priv.raw_addressing` value is only used for GQI_RDA, so we
remove it in favor of just checking against GQI_RDA
Signed-off-by: Bailey Forrest <bcf@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using `page_offset` like a boolean means a page may only be split into
two sections. With page sizes larger than 4k, this can be very wasteful.
Future commits in this patchset use `struct gve_rx_slot_page_info` in a
way which supports a fixed buffer size and a variable page size.
Signed-off-by: Bailey Forrest <bcf@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Future use cases will have a different padding value.
Signed-off-by: Bailey Forrest <bcf@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These functions will be shared by the GQI and DQO variants of the GVNIC
driver as of follow-up patches in this series.
Signed-off-by: Bailey Forrest <bcf@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch lets the driver reuse buffers that have been freed by the
networking stack.
In the raw addressing case, this allows the driver avoid allocating new
buffers.
In the qpl case, the driver can avoid copies.
This patch separates the page refcount tracking mechanism
into a function gve_rx_can_recycle_buffer which uses get_page - this will
be changed in a future patch to entirely eliminate the use of get_page in
tracking page refcounts.
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support to use raw dma addresses in the rx path. Due to this new
support we can alloc a new buffer instead of making a copy.
RX buffers are handed to the networking stack and are
re-allocated as needed, avoiding the need to use
skb_copy_to_linear_data() as in "qpl" mode.
Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current code use dma_wmb() to ensure Rx/Tx descriptors are visible
to device before writing to doorbell.
However, these dma_wmb() are wrong and unnecessary. Therefore,
they should be removed.
iowrite32be() called from gve_rx_write_doorbell()/gve_tx_put_doorbell()
should guaratee that all previous writes to WB/UC memory is visible to
device before the write done by iowrite32be().
E.g. On ARM64, iowrite32be() calls __iowmb() which expands to dma_wmb()
and only then calls __raw_writel().
Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Synces the DMA buffer properly in order for CPU and device to see
the most up-to-data data.
Signed-off-by: Yangchun Fu <yangchun@google.com>
Reviewed-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
desc_cnt and data_cnt should always be equal. In the case of a dropped
packet desc_cnt was still getting updated (correctly), data_cnt
was not. To eliminate this bug and prevent it from recurring this
patch combines them into one ring level cnt.
Signed-off-by: Catherine Sullivan <csully@google.com>
Reviewed-by: Sagi Shahar <sagis@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Variables allocated by kvzalloc should not be freed by kfree.
Because they may be allocated by vmalloc.
So we replace kfree with kvfree here.
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On unusual page sizes, we get harmless warnings:
drivers/net/ethernet/google/gve/gve_rx.c:283:6: error: unused variable 'pagecount' [-Werror,-Wunused-variable]
drivers/net/ethernet/google/gve/gve_rx.c:336:1: error: unused label 'have_skb' [-Werror,-Wunused-label]
Change the preprocessor #if to regular if() to avoid this.
Fixes: f5cedc84a3 ("gve: Add transmit and receive support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for the workqueue to handle management interrupts and
support for resets.
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: Sagi Shahar <sagis@google.com>
Signed-off-by: Jon Olson <jonolson@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Luigi Rizzo <lrizzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for passing traffic.
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: Sagi Shahar <sagis@google.com>
Signed-off-by: Jon Olson <jonolson@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Luigi Rizzo <lrizzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
