When using XDP every ingress packet is passed to an eBPF (xdp) program
which returns an action for this packet.
This patch adds counters for the number of times each such action was
received. It also counts all the invalid actions received from the eBPF
program.
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The new metrics provide granular visibility along multiple network
dimensions and enable troubleshooting and remediation of issues caused
by instances exceeding network performance allowances.
The new statistics can be queried using ethtool command.
Signed-off-by: Guy Tzalik <gtzalik@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most statistics in ena driver are incremented, meaning that a stat's
value is a sum of all increases done to it since driver/queue
initialization.
This patch makes all statistics this way, effectively making missed_tx
statistic incremental.
Also added a comment regarding rx_drops and tx_drops to make it
clearer how these counters are calculated.
Fixes: 11095fdb71 ("net: ena: add statistics for missed tx packets")
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ena_del_napi_in_range() function unregisters the napi handler for
rings in a given range.
This function had the following WARN_ON macro:
WARN_ON(ENA_IS_XDP_INDEX(adapter, i) &&
adapter->ena_napi[i].xdp_ring);
This macro prints the call stack if the expression inside of it is
true [1], but the expression inside of it is the wanted situation.
The expression checks whether the ring has an XDP queue and its index
corresponds to a XDP one.
This patch changes the expression to
!ENA_IS_XDP_INDEX(adapter, i) && adapter->ena_napi[i].xdp_ring
which indicates an unwanted situation.
Also, change the structure of the function. The napi handler is
unregistered for all rings, and so there's no need to check whether the
index is an XDP index or not. By removing this check the code becomes
much more readable.
Fixes: 548c4940b9 ("net: ena: Implement XDP_TX action")
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The reset work is scheduled by the timer routine whenever it
detects that a device reset is required (e.g. when a keep_alive signal
is missing).
When releasing device resources in ena_destroy_device() the driver
cancels the scheduling of the timer routine without destroying the reset
work explicitly.
This creates the following bug:
The driver is suspended and the ena_suspend() function is called
-> This function calls ena_destroy_device() to free the net device
resources
-> The driver waits for the timer routine to finish
its execution and then cancels it, thus preventing from it
to be called again.
If, in its final execution, the timer routine schedules a reset,
the reset routine might be called afterwards,and a redundant call to
ena_restore_device() would be made.
By changing the reset routine we allow it to read the device's state
accurately.
This is achieved by checking whether ENA_FLAG_TRIGGER_RESET flag is set
before resetting the device and making both the destruction function and
the flag check are under rtnl lock.
The ENA_FLAG_TRIGGER_RESET is cleared at the end of the destruction
routine. Also surround the flag check with 'likely' because
we expect that the reset routine would be called only when
ENA_FLAG_TRIGGER_RESET flag is set.
The destruction of the timer and reset services in __ena_shutoff() have to
stay, even though the timer routine is destroyed in ena_destroy_device().
This is to avoid a case in which the reset routine is scheduled after
free_netdev() in __ena_shutoff(), which would create an access to freed
memory in adapter->flags.
Fixes: 8c5c7abdeb ("net: ena: add power management ops to the ENA driver")
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that BPF program/link management is centralized in generic net_device
code, kernel code never queries program id from drivers, so
XDP_QUERY_PROG/XDP_QUERY_PROG_HW commands are unnecessary.
This patch removes all the implementations of those commands in kernel, along
the xdp_attachment_query().
This patch was compile-tested on allyesconfig.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200722064603.3350758-10-andriin@fb.com
New devices add a new hardware acceleration engine, which adds some
restrictions to the driver.
Metadata descriptor must be present for each packet and the maximum
burst size between two doorbells is now limited to a number
advertised by the device.
This patch adds:
1. A handshake protocol between the driver and the device, so the
device will enable the accelerated queues only when both sides
support it.
2. The driver support for the new acceleration engine:
2.1. Send metadata descriptor for each Tx packet.
2.2. Limit the number of packets sent between doorbells.(*)
(*) A previous driver implementation of this feature was comitted in
commit 05d62ca218 ("net: ena: add handling of llq max tx burst size")
however the design of the interface between the driver and device
changed since then. This change is reflected in this commit.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the ENA device resets to recover from some error state, all LLQ
configuration values are reset to their defaults, because LLQ is
initialized only once during ena_probe().
Changes in this commit:
1. Move the LLQ configuration process into ena_init_device()
which is called from both ena_probe() and ena_restore_device(). This
way, LLQ setup configurations that are different from the default
values will survive resets.
2. Extract the LLQ bar mapping to ena_map_llq_bar(),
and call once in the lifetime of the driver from ena_probe(),
since there is no need to unmap and map the LLQ bar again every reset.
3. Map the LLQ bar if it exists, regardless if initialization of LLQ
placement policy (ENA_ADMIN_PLACEMENT_POLICY_DEV) succeeded
or not. Initialization might fail the first time, falling back to the
ENA_ADMIN_PLACEMENT_POLICY_HOST placement policy, but later succeed
after device reset, in which case the LLQ bar needs to be mapped
already.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the rss_configurable_function_key bit to driver_supported_feature.
This bit tells the device that the driver in question supports the
retrieving and updating of RSS function and hash key, and therefore
the device should allow RSS function and key manipulation.
This commit turns on device support for hash key and RSS function
management. Without this commit this feature is turned off at the
device and appears to the user as unsupported.
This commit concludes the following series of already merged commits:
commit 0af3c4e2ea ("net: ena: changes to RSS hash key allocation")
commit c1bd17e51c ("net: ena: change default RSS hash function to Toeplitz")
commit f66c2ea3b1 ("net: ena: allow setting the hash function without changing the key")
commit e9a1de378d ("net: ena: fix error returning in ena_com_get_hash_function()")
commit 80f8443fcd ("net: ena: avoid unnecessary admin command when RSS function set fails")
commit 6a4f7dc82d ("net: ena: rss: do not allocate key when not supported")
commit 0d1c3de7b8 ("net: ena: fix incorrect default RSS key")
The above commits represent the last part of the implementation of
this feature, and with them merged the feature can be enabled
in the device.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for traffic mirroring, where the hardware reads the
buffer from the instance memory directly.
Traffic Mirroring needs access to the rx buffers in the instance.
To have this access, this patch:
1. Changes the code to map and unmap the rx buffers bidirectionally.
2. Enables the relevant bit in driver_supported_features to indicate
to the FW that this driver supports traffic mirroring.
Rx completion is not generated until mirroring is done to avoid
the situation where the driver changes the buffer before it is
mirrored.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
gcc 4.8 reports a warning when initializing with = {0}.
Dropping the "0" from the braces fixes the issue.
This fix is not ANSI compatible but is allowed by gcc.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For an overview of the race created by this patch goto synchronization
label.
In napi busy-poll mode, the kernel invokes the napi handler of the
device repeatedly to poll the NIC's receive queues. This process
repeats until a timeout, specific for each connection, is up.
By polling packets in busy-poll mode the user may gain lower latency
and higher throughput (since the kernel no longer waits for interrupts
to poll the queues) in expense of CPU usage.
Upon completing a napi routine, the driver checks whether
the routine was called by an interrupt handler. If so, the driver
re-enables interrupts for the device. This is needed since an
interrupt routine invocation disables future invocations until
explicitly re-enabled.
The driver avoids re-enabling the interrupts if they were not disabled
in the first place (e.g. if driver in busy mode).
Originally, the driver checked whether interrupt re-enabling is needed
by reading the 'ena_napi->unmask_interrupt' variable. This atomic
variable was set upon interrupt and cleared after re-enabling it.
In the 4.10 Linux version, the 'napi_complete_done' call was changed
so that it returns 'false' when device should not re-enable
interrupts, and 'true' otherwise. The change includes reading the
"NAPIF_STATE_IN_BUSY_POLL" flag to check if the napi call is in
busy-poll mode, and if so, return 'false'.
The driver was changed to re-enable interrupts according to this
routine's return value.
The Linux community rejected the use of the
'ena_napi->unmaunmask_interrupt' variable to determine whether
unmasking is needed, and urged to use napi_napi_complete_done()
return value solely.
See https://lore.kernel.org/patchwork/patch/741149/ for more details
As explained, a busy-poll session exists for a specified timeout
value, after which it exits the busy-poll mode and re-enters it later.
This leads to many invocations of the napi handler where
napi_complete_done() false indicates that interrupts should be
re-enabled.
This creates a bug in which the interrupts are re-enabled
unnecessarily.
To reproduce this bug:
1) echo 50 | sudo tee /proc/sys/net/core/busy_poll
2) echo 50 | sudo tee /proc/sys/net/core/busy_read
3) Add counters that check whether
'ena_unmask_interrupt(tx_ring, rx_ring);'
is called without disabling the interrupts in the first
place (i.e. with calling the interrupt routine
ena_intr_msix_io())
Steps 1+2 enable busy-poll as the default mode for new connections.
The busy poll routine rearms the interrupts after every session by
design, and so we need to add an extra check that the interrupts were
masked in the first place.
synchronization:
This patch introduces a race between the interrupt handler
ena_intr_msix_io() and the napi routine ena_io_poll().
Some macros and instruction were added to prevent this race from leaving
the interrupts masked. The following specifies the different race
scenarios in this patch:
1) interrupt handler and napi routine run sequentially
i) interrupt handler is called, sets 'interrupts_masked' flag and
successfully schedules the napi handler via softirq.
In this scenario the napi routine might not see the flag change
for several reasons:
a) The flag is stored in a register by the compiler. For this
case the WRITE_ONCE macro which prevents this.
b) The compiler might reorder the instruction. For this the
smp_wmb() instruction was used which implies a compiler memory
barrier.
c) On archs with weak consistency model (like ARM64) the napi
routine might be scheduled and start running before the flag
STORE instruction is committed to cache/memory. To ensure this
doesn't happen, the smp_wmb() instruction was added. It ensures
that the flag set instruction is committed before scheduling
napi.
ii) compiler reorders the flag's value check in the 'if' with
the flag set in the napi routine.
This scenario is prevented by smp_rmb() call after the flag check.
2) interrupt handler and napi routine run in parallel (can happen when
busy poll routine invokes the napi handler)
i) interrupt handler sets the flag in one core, while the napi
routine reads it in another core.
This scenario also is divided into two cases:
a) napi_complete_done() doesn't finish running, in which case
napi_sched() would just set NAPIF_STATE_MISSED and the napi
routine would reschedule itself without changing the flag's value.
b) napi_complete_done() finishes running. In this case the
napi routine might override the flag's value.
This doesn't present any rise since it later unmasks the
interrupt vector.
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix sparse build warning:
drivers/net/ethernet/amazon/ena/ena_netdev.c:2193:34: warning:
Using plain integer as NULL pointer
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Suggested-by: Joe Perches <joe@perches.com>
Acked-by: Shay Agroskin <shayagr@amazon.com>
Acked-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With legacy PM, drivers themselves were responsible for managing the
device's power states and takes care of register states.
After upgrading to the generic structure, PCI core will take care of
required tasks and drivers should do only device-specific operations.
Compile-tested only.
Signed-off-by: Vaibhav Gupta <vaibhavgupta40@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes two issues with XDP:
1. If the XDP verdict is XDP_ABORTED we break the loop, which results in
us handling one buffer per napi cycle instead of the total budget
(usually 64). To overcome this simply change the xdp_verdict check to
!= XDP_PASS. When the verdict is XDP_PASS, the skb is not expected to
be NULL.
2. Update the residual budget for XDP_DROP and XDP_ABORTED, since
packets are handled in these cases.
Fixes: 548c4940b9 ("net: ena: Implement XDP_TX action")
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When sending very high packet rate, the XDP tx queues can get full and
start dropping packets. In this case we don't free the pages which
results in ena driver draining the system memory.
Fix:
Simply free the pages when necessary.
Fixes: 548c4940b9 ("net: ena: Implement XDP_TX action")
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to use standard 'xdp' prefix, rename convert_to_xdp_frame
utility routine in xdp_convert_buff_to_frame and replace all the
occurrences
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/bpf/6344f739be0d1a08ab2b9607584c4d5478c8c083.1590698295.git.lorenzo@kernel.org
This commit reduces the driver load time by using usec resolution
instead of msec when polling for hardware state change.
Also add back-off mechanism to handle cases where minimal sleep
time is not enough.
Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. Reorder sanity checks in get_comp_ctxt() to make more sense
2. Reorder variables in ena_com_fill_hash_function() and
ena_calc_io_queue_size() in reverse christmas tree.
3. Move around member initializations.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. Join unnecessarily broken short lines in ena_com.c ena_netdev.c
2. Fix Indentations of broken lines
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
fix spelling and grammar mistakes in comments in ena_com.h,
ena_com.c and ena_netdev.c
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make all types of variables that convey the number and sizeof queues to
be u32, for consistency with the API between the driver and device via
ena_admin_defs.h:ena_admin_get_feat_resp.max_queue_ext fields. Current
code sometimes uses int and there are multiple assignments between these
variables with different types.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename ena_com_free_desc to ena_com_free_q_entries to match
the LLQ mode.
In non-LLQ mode, an entry in an IO ring corresponds to a
a descriptor. In LLQ mode an entry may correspond to several
descriptors (per LLQ definition).
Signed-off-by: Igor Chauskin <igorch@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Newer ENA devices can write data to rx buffers with an offset
from the beginning of the buffer.
This commit adds support for this feature in the driver.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Frame size ENA_PAGE_SIZE is limited to 16K on systems with larger
PAGE_SIZE than 16K. Change ENA_XDP_MAX_MTU to also take into account
the reserved tailroom.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Sameeh Jubran <sameehj@amazon.com>
Cc: Arthur Kiyanovski <akiyano@amazon.com>
Link: https://lore.kernel.org/bpf/158945341384.97035.907403694833419456.stgit@firesoul
The 'ENA_REGS_RESET_SHUTDOWN' enum indicates a normal driver
shutdown / removal procedure.
Also, a comment is added to one of the reset reason assignments for
code clarity.
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. Add support for getting tx drops from the device and saving them
in the driver.
2. Report tx via netdev stats.
Signed-off-by: Igor Chauskin <igorch@amazon.com>
Signed-off-by: Guy Tzalik <gtzalik@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add unmask interrupts statistics to ethtool.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently in the driver we are setting the hash function to be CRC32.
Starting with this commit we want to change the default behaviour so that
we set the hash function to be Toeplitz instead.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix sparse warnings:
drivers/net/ethernet/amazon/ena/ena_netdev.c:460:6: warning: symbol 'ena_xdp_exchange_program_rx_in_range' was not declared. Should it be static?
drivers/net/ethernet/amazon/ena/ena_netdev.c:481:6: warning: symbol 'ena_xdp_exchange_program' was not declared. Should it be static?
drivers/net/ethernet/amazon/ena/ena_netdev.c:1555:5: warning: symbol 'ena_xdp_handle_buff' was not declared. Should it be static?
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Overlapping header include additions in macsec.c
A bug fix in 'net' overlapping with the removal of 'version'
string in ena_netdev.c
Overlapping test additions in selftests Makefile
Overlapping PCI ID table adjustments in iwlwifi driver.
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently ENA only provides the PCI remove() handler, used during rmmod
for example. This is not called on shutdown/kexec path; we are potentially
creating a failure scenario on kexec:
(a) Kexec is triggered, no shutdown() / remove() handler is called for ENA;
instead pci_device_shutdown() clears the master bit of the PCI device,
stopping all DMA transactions;
(b) Kexec reboot happens and the device gets enabled again, likely having
its FW with that DMA transaction buffered; then it may trigger the (now
invalid) memory operation in the new kernel, corrupting kernel memory area.
This patch aims to prevent this, by implementing a shutdown() handler
quite similar to the remove() one - the difference being the handling
of the netdev, which is unregistered on remove(), but following the
convention observed in other drivers, it's only detached on shutdown().
This prevents an odd issue in AWS Nitro instances, in which after the 2nd
kexec the next one will fail with an initrd corruption, caused by a wild
DMA write to invalid kernel memory. The lspci output for the adapter
present in my instance is:
00:05.0 Ethernet controller [0200]: Amazon.com, Inc. Elastic Network
Adapter (ENA) [1d0f:ec20]
Suggested-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com>
Acked-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
last_keep_alive_jiffies is updated in probe and when a keep-alive
event is received. In case the driver times-out on a keep-alive event,
it has high chances of continuously timing-out on keep-alive events.
This is because when the driver recovers from the keep-alive-timeout reset
the value of last_keep_alive_jiffies is very old, and if a keep-alive
event is not received before the next timer expires, the value of
last_keep_alive_jiffies will cause another keep-alive-timeout reset
and so forth in a loop.
Solution:
Update last_keep_alive_jiffies whenever the device is restored after
reset.
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Noam Dagan <ndagan@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rx req_id is an index in struct ena_eth_io_rx_cdesc_base.
The driver should validate that the Rx req_id it received from
the device is in range [0, ring_size -1]. Failure to do so could
yield to potential memory access violoation.
The validation was mistakenly done when refilling
the Rx submission queue and not in Rx completion queue.
Fixes: ad974baef2 ("net: ena: add support for out of order rx buffers refill")
Signed-off-by: Noam Dagan <ndagan@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bug:
In short the main issue is caused by the fact that the number of queues
is changed using ethtool after ena_probe() has been called and before
ena_up() was executed. Here is the full scenario in detail:
* ena_probe() is called when the driver is loaded, the driver is not up
yet at the end of ena_probe().
* The number of queues is changed -> io_queue_count is changed as well -
ena_up() is not called since the "dev_was_up" boolean in
ena_update_queue_count() is false.
* ena_up() is called by the kernel (it's called asynchronously some
time after ena_probe()). ena_setup_io_intr() is called by ena_up() and
it uses io_queue_count to get the suitable irq lines for each msix
vector. The function ena_request_io_irq() is called right after that
and it uses msix_vecs - This value only changes during ena_probe() and
ena_restore() - to request the irq vectors. This results in "Failed to
request I/O IRQ" error for i > io_queue_count.
Numeric example:
* After ena_probe() io_queue_count = 8, msix_vecs = 9.
* The number of queues changes to 4 -> io_queue_count = 4, msix_vecs = 9.
* ena_up() is executed for the first time:
** ena_setup_io_intr() inits the vectors only up to io_queue_count.
** ena_request_io_irq() calls request_irq() and fails for i = 5.
How to reproduce:
simply run the following commands:
sudo rmmod ena && sudo insmod ena.ko;
sudo ethtool -L eth1 combined 3;
Fix:
Use ENA_MAX_MSIX_VEC(adapter->num_io_queues + adapter->xdp_num_queues)
instead of adapter->msix_vecs. We need to take XDP queues into
consideration as they need to have msix vectors assigned to them as well.
Note that the XDP cannot be attached before the driver is up and running
but in XDP mode the issue might occur when the number of queues changes
right after a reset trigger.
The ENA_MAX_MSIX_VEC simply adds one to the argument since the first msix
vector is reserved for management queue.
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Overview:
We don't frequently change the msix vectors throughout the life cycle of
the driver. We do so in two functions: ena_probe() and ena_restore().
ena_probe() is only called when the driver is loaded. ena_restore() on the
other hand is called during device reset / resume operations.
We use num_io_queues for calculating and allocating the number of msix
vectors. At ena_probe() this value is equal to max_num_io_queues and thus
this is not an issue, however ena_restore() might be called after the
number of io queues has changed.
A possible bug scenario is as follows:
* Change number of queues from 8 to 4.
(num_io_queues = 4, max_num_io_queues = 8, msix_vecs = 9,)
* Trigger reset occurs -> ena_restore is called.
(num_io_queues = 4, max_num_io_queues =8 , msix_vecs = 5)
* Change number of queues from 4 to 6.
(num_io_queues = 6, max_num_io_queues = 8, msix_vecs = 5)
* The driver will reset due to failure of check_for_rx_interrupt_queue()
Fix:
This can be easily fixed by always using max_num_io_queues to init the
msix_vecs, since this number won't change as opposed to num_io_queues.
Fixes: 4d19266022 ("net: ena: multiple queue creation related cleanups")
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In this commit we revert the part of
commit 1a63443afd ("net/amazon: Ensure that driver version is aligned to the linux kernel"),
which breaks the interface between the ENA driver and FW.
We also replace the use of DRIVER_VERSION with DRIVER_GENERATION
when we bring back the deleted constants that are used in interface with
ENA device FW.
This commit does not change the driver version reported to the user via
ethtool, which remains the kernel version.
Fixes: 1a63443afd ("net/amazon: Ensure that driver version is aligned to the linux kernel")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Upstream drivers are managed inside global repository and released all
together, this ensure that driver version is the same as linux kernel,
so update amazon drivers to properly reflect it.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
>From the documentation of round_jiffies():
"Rounds a time delta in the future (in jiffies) up or down to
(approximately) full seconds. This is useful for timers for which
the exact time they fire does not matter too much, as long as
they fire approximately every X seconds.
By rounding these timers to whole seconds, all such timers will fire
at the same time, rather than at various times spread out. The goal
of this is to have the CPU wake up less, which saves power."
There are 2 parts to this patch:
================================
Part 1:
-------
In our case we need timer_service to be called approximately every
X=1 seconds, and the exact time does not matter, so using round_jiffies()
is the right way to go.
Therefore we add round_jiffies() to the mod_timer() in ena_timer_service().
Part 2:
-------
round_jiffies() is used in check_for_missing_keep_alive() when
getting the jiffies of the expiration of the keep_alive timeout. Here it
is actually a mistake to use round_jiffies() because we want the exact
time when keep_alive should expire and not an approximate rounded time,
which can cause early, false positive, timeouts.
Therefore we remove round_jiffies() in the calculation of
keep_alive_expired() in check_for_missing_keep_alive().
Fixes: 82ef30f13b ("net: ena: add hardware hints capability to the driver")
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/amazon/ena/ena_netdev.c: In function ena_xdp_xmit_buff:
drivers/net/ethernet/amazon/ena/ena_netdev.c:316:19: warning:
variable rx_ring set but not used [-Wunused-but-set-variable]
commit 548c4940b9 ("net: ena: Implement XDP_TX action")
left behind this unused variable.
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The first_interrupt field is accessed in ena_intr_msix_io() upon
receiving an interrupt.The rx_ring and tx_ring fields of napi can
be NULL when receiving interrupt for xdp queues. This patch fixes
the issue by moving the field to the ena_napi struct.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit implements the XDP_TX action in the ena driver. We allocate
separate tx queues for the XDP_TX. We currently allow xdp only when
there is enough queues to allocate for xdp.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit implements the basic functionality of drop/pass logic in the
ena driver.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In netpoll the napi handler could be called with budget equal to zero.
Current ENA napi handler doesn't take that into consideration.
The napi handler handles Rx packets in a do-while loop.
Currently, the budget check happens only after decrementing the
budget, therefore the napi handler, in rare cases, could run over
MAX_INT packets.
In addition to that, this moves all budget related variables to int
calculation and stop mixing u32 to avoid ambiguity
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set channels callback enables the user to change the count of queues
used by the driver using ethtool. We decided to currently support only
equal number of rx and tx queues, this might change in the future.
Also rename dev_up to dev_was_up in ena_update_queue_count() to make
it clearer.
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The number of queues can be derived using ethtool, no need to print
it in ena_probe()
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- Update ena_ethtool:ena_get_channels() to return adapter->max_io_queues
so that ethtool -l returns the correct maximum queue number.
- Change the name of ena_calc_io_queue_num() to
ena_calc_max_io_queue_num() as it returns the maximum number of io
queues and actual number of queues can be smaller if changed
by ethtool -L which is implemented in a later commit.
- Change variable name from io_queue_num to max_num_io_queues in
ena_calc_max_io_queue_num() and ena_probe().
- Make all types of variables that convey the number and sizeof queues
to be u32, for consistency with the API between the driver and the
device.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- Rename ena_calc_queue_size() to ena_calc_io_queue_size() for clarity
and consistency
- Remove redundant number of io queues parameter in functions
ena_enable_msix() and ena_enable_msix_and_set_admin_interrupts(),
which already get adapter parameter, so use adapter->num_io_queues
in the function instead.
- Use the local variable ena_dev instead of ctx->ena_dev in
ena_calc_io_queue_size
- Fix multi row comment alignments
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most places in the code refer to the IO queues as io_queues and not
simply queues. Examples - max_io_queues_per_vf, ENA_MAX_NUM_IO_QUEUES,
ena_destroy_all_io_queues() etc..
We are also adding the new max_num_io_queues field to struct ena_adapter
in the following commit.
The changes included in this commit are:
struct ena_adapter->num_queues => struct ena_adapter->num_io_queues
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ena_dev->intr_moder_rx/tx_interval save the intervals received from the
user after dividing them by ena_dev->intr_delay_resolution. Therefore
when intr_delay_resolution changes, the code needs to first mutiply
intr_moder_rx/tx_interval by the previous intr_delay_resolution to get
the value originally given by the user, and only then divide it by the
new intr_delay_resolution.
Current code does not first multiply intr_moder_rx/tx_interval by the old
intr_delay_resolution. This commit fixes it.
Also initialize ena_dev->intr_delay_resolution to be 1.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current implementation always updates the interrupt register with
the smoothed_interval of the rx_ring. However this should be
done only in case of adaptive interrupt moderation. If non-adaptive
interrupt moderation is used, the non-adaptive interrupt moderation
interval should be used. This commit fixes that.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Deleted unused 4 fields from struct ena_adapter and their only user
ena_restore_ethtool_params().
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1. Out of the fields {per_napi_bytes, per_napi_packets} in struct ena_ring,
only rx_ring->per_napi_packets are used to determine if napi did work
for dim.
This commit removes all other uses of these fields.
2. Remove ena_ring->moder_tbl_idx, which is not used by dim.
3. Remove all calls to ena_com_destroy_interrupt_moderation(), since all it
did was to destroy the interrupt moderation table, which is removed as
part of removing old interrupt moderation code.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add driver_supported_features to host_host info which is a new API used to
communicate to the device which features are supported by the driver.
Add the interrupt_moderation bit to host_info->driver_supported_features
and enable it to signal the device that this driver supports interrupt
moderation properly.
Reserved bits are for features implemented in the future
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the dim library for the rx adaptive interrupt moderation implementation
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add intr_moder_rx_interval to struct ena_com_dev and use it as the
location where the interrupt moderation rx interval is saved, instead
of the interrupt moderation table.
This is done as a first step before removing the old interrupt moderation
code.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a race condition that can occur when calling ena_down().
The ena_clean_tx_irq() - which is a part of the napi handler -
function might wake up the tx queue when the queue is supposed
to be down (during recovery or changing the size of the queues
for example) This causes the ena_start_xmit() function to trigger
and possibly try to access the destroyed queues.
The race is illustrated below:
Flow A: Flow B(napi handler)
ena_down()
netif_carrier_off()
netif_tx_disable()
ena_clean_tx_irq()
netif_tx_wake_queue()
ena_napi_disable_all()
ena_destroy_all_io_queues()
After these flows the tx queue is active and ena_start_xmit() accesses
the destroyed queue which leads to a kernel panic.
fixes: 1738cd3ed3 (net: ena: Add a driver for Amazon Elastic Network Adapters (ENA))
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 315c28d2b7 ("net: ena: ethtool: add extra properties retrieval via get_priv_flags").
As discussed at netconf and on the mailing list we can't allow
for the the abuse of private flags for exposing arbitrary device
labels.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current code of create_queues_with_size_backoff() allows the ring size
to become as small as ENA_MIN_RING_SIZE/2. This is a bug since we don't
want the queue ring to be smaller than ENA_MIN_RING_SIZE
In this commit we change the loop's termination condition to look at the
queue size of the next iteration instead of that of the current one,
so that the minimal queue size again becomes ENA_MIN_RING_SIZE.
Fixes: eece4d2ab9 ("net: ena: add ethtool function for changing io queue sizes")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let the compiler decide if the function should be inline in *.c files
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement the set_ringparam() function of the ethtool interface
to enable the changing of io queue sizes.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If there is not enough memory to allocate io queues the driver will
try to allocate smaller queues.
The backoff algorithm is as follows:
1. Try to allocate TX and RX and if successful.
1.1. return success
2. Divide by 2 the size of the larger of RX and TX queues (or both if their size is the same).
3. If TX or RX is smaller than 256
3.1. return failure.
4. else
4.1. go back to 1.
Also change the tx_queue_size, rx_queue_size field names in struct
adapter to requested_tx_queue_size and requested_rx_queue_size, and
use RX and TX queue 0 for actual queue sizes.
Explanation:
The original fields were useless as they were simply used to assign
values once from them to each of the queues in the adapter in ena_probe().
They could simply be deleted. However now that we have a backoff
feature, we have use for them. In case of backoff there is a difference
between the requested queue sizes and the actual sizes. Therefore there
is a need to save the requested queue size for future retries of queue
allocation (for example if allocation failed and then ifdown + ifup was
called we want to start the allocation from the original requested size of
the queues).
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently ethtool -g shows the same size for current and max queue
sizes.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use MAX_QUEUES_EXT get feature capability to query the device.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a new statistics to ETHTOOL to specify if the device calculated
and validated the Rx csum.
Signed-off-by: Evgeny Shmeilin <evgeny@annapurnaLabs.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reverse christmas tree arrangement is when strings are written from longer
to shorter with each line. Most of our functions are abiding this
arrangement but this function does not.
In this commit we arrange the variables of ena_probe() in reverse christmas
tree.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct ena_ring holds a union of free_rx_ids and free_tx_ids.
Both of the above fields mean the exact same thing and are used
exactly the same way.
Furthermore, these fields are always used with a prefix of the
type of ring. So for tx it will be tx_ring->free_tx_ids, and for
rx it will be rx_ring->free_rx_ids, which shows how redundant the
"_tx" and "_rx" parts are.
Furthermore still, this may lead to confusing code like where
tx_ring->free_rx_ids which works correctly but looks like a mess.
This commit removes the aforementioned redundancy by replacing the
free_rx/tx_ids union with a single free_ids field.
It also changes a single goto label name from err_free_tx_ids: to
err_tx_free_ids: for consistency with the above new notation.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit adds a mechanism for exposing different device
properties via ethtool's priv_flags. The strings are provided
by the device and copied to user space through the driver.
In this commit we:
Add commands, structs and defines necessary for handling
extra properties
Add functions for:
Allocation/destruction of a buffer for extra properties strings.
Retreival of extra properties strings and flags from the network device.
Handle the allocation of a buffer for extra properties strings.
* Initialize buffer with extra properties strings from the
network device at driver startup.
Use ethtool's get_priv_flags to expose extra properties of
the ENA device
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a maximum TX burst size that the ENA device can handle.
It is exposed by the device to the driver and the driver
needs to comply with it to avoid bugs.
In this commit we:
1. Add ena_com_is_doorbell_needed(), which calculates the number of
llq entries that will be used to hold a packet, and will return
true if they exceed the number of allowed entries in a burst.
If the function returns true, a doorbell needs to be invoked
to send this packet in the next burst.
2. Follow the available entries in the current burst:
- Every doorbell a new burst begins
- With each write of an llq entry, the available entries in the
current burst are decreased by 1.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
GCC 8 contains a number of new warnings as well as enhancements to existing
checkers. The warning - Wstringop-truncation - warns for calls to bounded
string manipulation functions such as strncat, strncpy, and stpncpy that
may either truncate the copied string or leave the destination unchanged.
In our case the destination string length (32 bytes) is much shorter than
the source string (64 bytes) which causes this warning to show up. In
general the destination has to be at least a byte larger than the length
of the source string with strncpy for this warning not to showup.
This can be easily fixed by using strlcpy instead which already does the
truncation to the string. Documentation for this function can be
found here:
https://elixir.bootlin.com/linux/latest/source/lib/string.c#L141
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ena_up() calls ena_init_napi() but does not call ena_del_napi() in
case of failure. This causes a segmentation fault upon rmmod when
netif_napi_del() is called. Fix this bug by calling ena_del_napi()
before returning error from ena_up().
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In some cases when a queue related allocation fails, successful past
allocations are freed but the pointer that pointed to them is not
set to NULL. This is a problem for 2 reasons:
1. This is generally a bad practice since this pointer might be
accidentally accessed in the future.
2. Future allocations using the same pointer check if the pointer
is NULL and fail if it is not.
Fixed this by setting such pointers to NULL in the allocation of
queue related objects.
Also refactored the code of ena_setup_tx_resources() to goto-style
error handling to avoid code duplication of resource freeing.
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are two reasons for this.
First, the xmit_more flag conceptually doesn't fit into the skb, as
xmit_more is not a property related to the skb.
Its only a hint to the driver that the stack is about to transmit another
packet immediately.
Second, it was only done this way to not have to pass another argument
to ndo_start_xmit().
We can place xmit_more in the softnet data, next to the device recursion.
The recursion counter is already written to on each transmit. The "more"
indicator is placed right next to it.
Drivers can use the netdev_xmit_more() helper instead of skb->xmit_more
to check the "more packets coming" hint.
skb->xmit_more is retained (but always 0) to not cause build breakage.
This change takes care of the simple s/skb->xmit_more/netdev_xmit_more()/
conversions. Remaining drivers are converted in the next patches.
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
After the previous patch, all the callers of ndo_select_queue()
provide as a 'fallback' argument netdev_pick_tx.
The only exceptions are nested calls to ndo_select_queue(),
which pass down the 'fallback' available in the current scope
- still netdev_pick_tx.
We can drop such argument and replace fallback() invocation with
netdev_pick_tx(). This avoids an indirect call per xmit packet
in some scenarios (TCP syn, UDP unconnected, XDP generic, pktgen)
with device drivers implementing such ndo. It also clean the code
a bit.
Tested with ixgbe and CONFIG_FCOE=m
With pktgen using queue xmit:
threads vanilla patched
(kpps) (kpps)
1 2334 2428
2 4166 4278
4 7895 8100
v1 -> v2:
- rebased after helper's name change
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix race condition between ena_update_on_link_change() and
ena_restore_device().
This race can occur if link notification arrives while the driver
is performing a reset sequence. In this case link can be set up,
enabling the device, before it is fully restored. If packets are
sent at this time, the driver might access uninitialized data
structures, causing kernel crash.
Move the clearing of ENA_FLAG_ONGOING_RESET and netif_carrier_on()
after ena_up() to ensure the device is ready when link is set up.
Fixes: d18e4f6834 ("net: ena: fix race condition between device reset and link up setup")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In ena_remove() we have the following stack call:
ena_remove()
unregister_netdev()
ena_destroy_device()
netif_carrier_off()
Calling netif_carrier_off() causes linkwatch to try to handle the
link change event on the already unregistered netdev, which leads
to a read from an unreadable memory address.
This patch switches the order of the two functions, so that
netif_carrier_off() is called on a regiestered netdev.
To accomplish this fix we also had to:
1. Remove the set bit ENA_FLAG_TRIGGER_RESET
2. Add a sanitiy check in ena_close()
both to prevent double device reset (when calling unregister_netdev()
ena_close is called, but the device was already deleted in
ena_destroy_device()).
3. Set the admin_queue running state to false to avoid using it after
device was reset (for example when calling ena_destroy_all_io_queues()
right after ena_com_dev_reset() in ena_down)
Fixes: 944b28aa29 ("net: ena: fix missing lock during device destruction")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During resume from hibernation if ena_restore_device fails,
ena_com_dev_reset() is called, and uses the readless read mechanism,
which was already destroyed by the call to
ena_com_mmio_reg_read_request_destroy(). This causes a NULL pointer
reference.
In this commit we switch the call order of the above two functions
to avoid this crash.
Fixes: d7703ddbd7 ("net: ena: fix rare bug when failed restart/resume is followed by driver removal")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts were easy to resolve using immediate context mostly,
except the cls_u32.c one where I simply too the entire HEAD
chunk.
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove redundant spinlock acquire parameter from ena_com_admin_init()
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently Rx refill is done when the number of required descriptors is
above 1/8 queue size. With a default of 1024 entries per queue the
threshold is 128 descriptors.
There is intention to increase the queue size to 8196 entries.
In this case threshold of 1024 descriptors is too large and can hurt
latency.
Add another limitation to Rx threshold to be at most 256 descriptors.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set skb->ip_summed to the correct value as reported by the device.
Add counter for the case where rx csum offload is enabled but
device didn't check it.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch includes all code changes necessary in ena_netdev to enable
packet sending via the LLQ placemnt mode.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces APIs for detection, initialization, configuration
and actual usage of low latency queues(LLQ). It extends transmit API with
creation of LLQ descriptors in device memory (which include host buffers
descriptors as well as packet header)
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Low Latency Queues(LLQ) allow usage of device's memory for descriptors
and headers. Such queues decrease processing time since data is already
located on the device when driver rings the doorbell.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new fields and definitions to host info and fill them
according to the latest ENA spec version.
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
napi poll functions should be initialized before running request_irq(),
to handle a rare condition where there is a pending interrupt, causing
the ISR to fire immediately while the poll function wasn't set yet,
causing a NULL dereference.
Fixes: 1738cd3ed3 ("net: ena: Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In a rare scenario when ena_device_restore() fails, followed by device
remove, an FLR will not be issued. In this case, the device will keep
sending asynchronous AENQ keep-alive events, even after driver removal,
leading to memory corruption.
Fixes: 8c5c7abdeb ("net: ena: add power management ops to the ENA driver")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Memory mapped with devm_ioremap is automatically freed when the driver
is disconnected from the device. Therefore there is no need to
explicitly call devm_iounmap.
Fixes: 0857d92f71 ("net: ena: add missing unmap bars on device removal")
Fixes: 411838e7b4 ("net: ena: fix rare kernel crash when bar memory remap fails")
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Minor conflict in net/core/rtnetlink.c, David Ahern's bug fix in 'net'
overlapped the renaming of a netlink attribute in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
ena uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Netanel Belgazal <netanel@amazon.com>
Cc: Saeed Bishara <saeedb@amazon.com>
Cc: Zorik Machulsky <zorik@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
module.h already contained moduleparam.h, so it is safe to remove
the redundant include.
The issue is detected with the help of Coccinelle.
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added memory barriers where they were missing to support multiple
architectures, and removed redundant ones.
As part of removing the redundant memory barriers and improving
performance, we moved to more relaxed versions of memory barriers,
as well as to the more relaxed version of writel - writel_relaxed,
while maintaining correctness.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
acquire the rtnl_lock during device destruction to avoid
using partially destroyed device.
ena_remove() shares almost the same logic as ena_destroy_device(),
so use ena_destroy_device() and avoid duplications.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
