Russell King says:
====================
net: mvneta: reduce size of TSO header allocation
With reference to
https://forum.turris.cz/t/random-kernel-exceptions-on-hbl-tos-7-0/18865/https://github.com/openwrt/openwrt/pull/12375#issuecomment-1528842334
It appears that mvneta attempts an order-6 allocation for the TSO
header memory. While this succeeds early on in the system's life time,
trying order-6 allocations later can result in failure due to memory
fragmentation.
Firstly, the reason it's so large is that we take the number of
transmit descriptors, and allocate a TSO header buffer for each, and
each TSO header is 256 bytes. The driver uses a simple mechanism to
determine the address - it uses the transmit descriptor index as an
index into the TSO header memory.
(The first obvious question is: do there need to be this
many? Won't each TSO header always have at least one bit
of data to go with it? In other words, wouldn't the maximum
number of TSO headers that a ring could accept be the number
of ring entries divided by 2?)
There is no real need for this memory to be an order-6 allocation,
since nothing in hardware requires this buffer to be contiguous.
Therefore, this series splits this order-6 allocation up into 32
order-1 allocations (8k pages on 4k page platforms), each giving
32 TSO headers per page.
In order to do this, these patches:
1) fix a horrible transmit path error-cleanup bug - the existing
code unmaps from the first descriptor that was allocated at
interface bringup, not the first descriptor that the packet
is using, resulting in the wrong descriptors being unmapped.
2) since xdp support was added, we now have buf->type which indicates
what this transmit buffer contains. Use this to mark TSO header
buffers.
3) get rid of IS_TSO_HEADER(), instead using buf->type to determine
whether this transmit buffer needs to be DMA-unmapped.
4) move tso_build_hdr() into mvneta_tso_put_hdr() to keep all the
TSO header building code together.
5) split the TSO header allocation into chunks of order-1 pages.
This has now been tested by the Turris folk and has been found to fix
the allocation error.
====================
Link: https://lore.kernel.org/r/ZFtuhJOC03qpASt2@shell.armlinux.org.uk
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Now that we no longer need to check whether the DMA address is within
the TSO header DMA memory range for the queue, we can allocate the TSO
header DMA memory in chunks rather than one contiguous order-6 chunk,
which can stress the kernel's memory subsystems to allocate.
Instead, use order-1 (8k) allocations, which will result in 32 order-1
pages containing 32 TSO headers.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move tso_build_hdr() into mvneta_tso_put_hdr() so that all the TSO
header building code is in one place.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Now that we use a different buffer type for TSO headers, we can use
buf->type to determine whether the original buffer was DMA-mapped or
not. The rules are:
MVNETA_TYPE_XDP_TX - from a DMA pool, no unmap is required
MVNETA_TYPE_XDP_NDO - dma_map_single()'d
MVNETA_TYPE_SKB - normal skbuff, dma_map_single()'d
MVNETA_TYPE_TSO - from the TSO buffer area
This means we only need to call dma_unmap_single() on the XDP_NDO and
SKB types of buffer, and we no longer need the private IS_TSO_HEADER()
which relies on the TSO region being contiguously allocated.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Mark dma-mapped skbs and TSO buffers separately, so we can use
buf->type to identify their differences.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The transmit code assumes that the transmit descriptors that are used
begin with the first descriptor in the ring, but this may not be the
case. Fix this by providing a new function that dma-unmaps a range of
numbered descriptor entries, and use that to do the unmapping.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Currently the SYN RTO schedule follows an exponential backoff
scheme, which can be unnecessarily conservative in cases where
there are link failures. In such cases, it's better to
aggressively try to retransmit packets, so it takes routers
less time to find a repath with a working link.
We chose a default value for this sysctl of 4, to follow
the macOS and IOS backoff scheme of 1,1,1,1,1,2,4,8, ...
MacOS and IOS have used this backoff schedule for over
a decade, since before this 2009 IETF presentation
discussed the behavior:
https://www.ietf.org/proceedings/75/slides/tcpm-1.pdf
This commit makes the SYN RTO schedule start with a number of
linear backoffs given by the following sysctl:
* tcp_syn_linear_timeouts
This changes the SYN RTO scheme to be: init_rto_val for
tcp_syn_linear_timeouts, exp backoff starting at init_rto_val
For example if init_rto_val = 1 and tcp_syn_linear_timeouts = 2, our
backoff scheme would be: 1, 1, 1, 2, 4, 8, 16, ...
Signed-off-by: David Morley <morleyd@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Tested-by: David Morley <morleyd@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20230509180558.2541885-1-morleyd.kernel@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The 'sched' index value must be checked before accessing an element
of the 'sctp_sched_ops' array. Otherwise, it can lead to OOB access.
Note that it's harmless since the 'sched' parameter is checked before
calling 'sctp_sched_set_sched'.
Found by InfoTeCS on behalf of Linux Verification Center
(linuxtesting.org) with SVACE.
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Ilia.Gavrilov <Ilia.Gavrilov@infotecs.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now macsec on top of vlan can be offloaded to macsec offloading
devices so that VLAN tag is sent in clear text on wire i.e,
packet structure is DMAC|SMAC|VLAN|SECTAG. Offloading devices can
simply enable NETIF_F_HW_MACSEC feature in netdev->vlan_features for
this to work. But the logic in offloading drivers to retrieve the
private structure from netdev needs to be changed to check whether
the netdev received is real device or a vlan device and get private
structure accordingly. This patch changes the offloading drivers to
use helper macsec_netdev_priv instead of netdev_priv.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
"ip link set dev "$devbond1" nomaster"
This line code in bond-eth-type-change.sh is unnecessary.
Because $devbond1 was not added to any master device.
Signed-off-by: Liang Li <liali@redhat.com>
Acked-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since veth is very likely to be enabled and there are some drivers
(e.g. mlx5) where CONFIG_PAGE_POOL_STATS is optional, make
CONFIG_PAGE_POOL_STATS optional for veth too in order to keep it
optional instead of required.
Suggested-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Horatiu Vultur says:
====================
net: lan966x: Add support for ES0 VCAP
Provide the Egress Stage 0 (ES0) VCAP (Versatile Content-Aware
Processor) support for the lan966x platform.
The ES0 VCAP has only 1 lookup which is accessible with a TC chain
id 10000000.
Currently only one action is support which is vlan pop. Also it is
possible to link the IS1 to ES0 using 'goto chain 10000000'.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Enable the TC command to use the lan966x ES0 VCAP. Currently support
only one action which is vlan pop, other will be added later.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add ES0 VCAP port keyset configuration for lan966x and also update
debugfs to show the keyset configuration.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provide ES0 (egress stage 0) VCAP model for lan966x.
This provides rewriting functionality in the gress path.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some DHCP server implementations only send the important requested DHCP
options in the final BOOTP reply (DHCPACK).
One example is systemd-networkd.
However, RFC2131, in section 4.3.1 states:
> The server MUST return to the client:
> [...]
> o Parameters requested by the client, according to the following
> rules:
>
> -- IF the server has been explicitly configured with a default
> value for the parameter, the server MUST include that value
> in an appropriate option in the 'option' field, ELSE
I've reported the issue here:
https://github.com/systemd/systemd/issues/27471
Linux PNP DHCP client implementation only takes into account the DNS
servers received in the first BOOTP reply (DHCPOFFER).
This usually isn't an issue as servers are required to put the same
values in the DHCPOFFER and DHCPACK.
However, RFC2131, in section 4.3.2 states:
> Any configuration parameters in the DHCPACK message SHOULD NOT
> conflict with those in the earlier DHCPOFFER message to which the
> client is responding. The client SHOULD use the parameters in the
> DHCPACK message for configuration.
When making Linux PNP DHCP client (cmdline ip=dhcp) interact with
systemd-networkd DHCP server, an interesting "protocol misunderstanding"
happens:
Because DNS servers were only specified in the DHCPACK and not in the
DHCPOFFER, Linux will not catch the correct DNS servers: in the first
BOOTP reply (DHCPOFFER), it sees that there is no DNS, and sets as
fallback the IP of the DHCP server itself. When the second BOOTP reply
comes (DHCPACK), it's already too late: the kernel will not overwrite
the fallback setting it has set previously.
This patch makes the kernel overwrite its DNS fallback by DNS servers
specified in the DHCPACK if any.
Signed-off-by: Martin Wetterwald <martin@wetterwald.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Bhupesh Sharma <bhupesh.sharma@linaro.org>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Nobuhiro Iwamatsu <nobuhiro1.iwamatsu@toshiba.co.jp>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The remove callback is only ever called after .probe() returned
successfully. After that get_stmmac_bsp_priv() always return non-NULL.
Side note: The early exit would also be a bug because the return value
of qcom_ethqos_remove() is ignored by the device core and the device is
unbound unconditionally. So exiting early resulted in a dangerous
resource leak as all devm allocated resources (some memory and the
register mappings) are freed but the network device stays around. Using
the network device afterwards probably oopses.
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The function returns zero unconditionally. Change it to return void instead
which simplifies some callers as error handing becomes unnecessary.
The function is also used for some drivers as remove callback. Switch these
to the .remove_new() callback. For some others no error can happen in the
remove callback now, convert them to .remove_new(), too.
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Xuan Zhuo says:
====================
virtio_net: refactor xdp codes
Due to historical reasons, the implementation of XDP in virtio-net
is relatively chaotic. For example, the processing of XDP actions
has two copies of similar code. Such as page, xdp_page processing, etc.
The purpose of this patch set is to refactor these code. Reduce the difficulty
of subsequent maintenance. Subsequent developers will not introduce new bugs
because of some complex logical relationships.
In addition, the supporting to AF_XDP that I want to submit later will
also need to reuse the logic of XDP, such as the processing of actions,
I don't want to introduce a new similar code. In this way, I can reuse
these codes in the future.
====================
Link: https://lore.kernel.org/r/20230508061417.65297-1-xuanzhuo@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Because the skb build code is not shared between xdp and non-xdp, and
the xdp code in receive_small() is simpler, so "skip_xdp" is not needed.
We can remove it.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Avoid the problem that some variables(headroom and so on) will repeat
the calculation when process xdp.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In the case of XDP-PASS, skb_reserve uses the "delta" to compatible
non-XDP, now that is not shared between xdp and non-xdp, so we can
remove this logic.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The purpose of this patch is to simplify the receive_small().
Separate all the logic of XDP of small into a function.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The purpose of this patch is to simplify the receive_mergeable().
Separate all the logic of XDP into a function.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
virtnet_build_xdp_buff_mrg() auto release xdp shinfo then the caller no
need to careful the xdp shinfo.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch introduce a new function that frees the rest mergeable buf.
The subsequent patch will reuse this function.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch introduce a new function that releases the
xdp shinfo. The subsequent patch will reuse this function.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
At present, we have two similar logic to perform the XDP prog.
Therefore, this patch separates the code of executing XDP, which is
conducive to later maintenance.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The previous patch, in order to facilitate review, I do not do any
modification. This patch has made some optimization on the top.
* remove some repeated logics in this function.
* add fast check for passing without any alloc.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Separating the logic of preparation for xdp from receive_mergeable.
The purpose of this is to simplify the logic of execution of XDP.
The main logic here is that when headroom is insufficient, we need to
allocate a new page and calculate offset. It should be noted that if
there is new page, the variable page will refer to the new page.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In the xdp implementation of virtio-net mergeable, it always checks
whether two page is used and a page is selected to release. This is
complicated for the processing of action, and be careful.
In the entire process, we have such principles:
* If xdp_page is used (PASS, TX, Redirect), then we release the old
page.
* If it is a drop case, we will release two. The old page obtained from
buf is release inside err_xdp, and xdp_page needs be relased by us.
But in fact, when we allocate a new page, we can release the old page
immediately. Then just one is using, we just need to release the new
page for drop case. On the drop path, err_xdp will release the variable
"page", so we only need to let "page" point to the new xdp_page in
advance.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
