Currently when going to/from a NULL XDP program the driver uses
ionic_stop_queues_reconfig() and then ionic_start_queues_reconfig() in
order to re-register the xdp_rxq_info and re-init the queues. This is
fine until page_pool(s) are used in an upcoming patch.
In preparation for adding page_pool support make sure to completely
rebuild the queues when going to/from a NULL XDP program. Without this
change the call to mem_allocator_disconnect() never happens when going
to a NULL XDP program, which eventually results in
xdp_rxq_info_reg_mem_model() failing with -ENOSPC due to the mem_id_pool
ida having no remaining space.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://patch.msgid.link/20240906232623.39651-6-brett.creeley@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If the doorbell workaround isn't required for a certain
asic_type then there is no need to run the associated
code. Since newer FW versions are finally reporting their
asic_type we can use a flag to determine whether or not to
do the workaround.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240619003257.6138-9-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We only support (u16)-1 size for rx_copybreak, so we can reduce the
field size and move a couple other fields around to save a little
space in the ionic_lif struct.
Before:
/* size: 17440, cachelines: 273, members: 56 */
/* sum members: 17403, holes: 9, sum holes: 37 */
/* last cacheline: 32 bytes */
After:
/* size: 17424, cachelines: 273, members: 56 */
/* sum members: 17401, holes: 7, sum holes: 23 */
/* last cacheline: 16 bytes */
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240619003257.6138-8-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a work item for each queue that will be run on the queue's
preferred cpu and will schedule another napi. This napi is
run in case the device missed a doorbell and didn't process
a packet. This is a problem for the Elba asic that happens
very rarely.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240619003257.6138-6-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Instead of using the system's default workqueue, add a private
workqueue for the device to use for its little jobs. This is
to better support the new work items we will be adding in the
next patches for PF and VF specific jobs, without inundating
the system workqueue in a couple of customer cases where our
devices get scaled out to 100-200 VFs.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240619003257.6138-4-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Remove the timer-per-queue mechanics from the missed doorbell
check in preparation for the new missed doorbell fix.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240619003257.6138-2-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rearange a few fields for better cache use and to put the
flags field up into the first cacheline rather than the last.
struct ionic_qcq
Before: /* size: 2176, cachelines: 34, members: 23 */
After: /* size: 2112, cachelines: 33, members: 23 */
Reviewed-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
An earlier change moved the hwstamp queue check into a helper
function with an unlikely(). However, it makes more sense for
the caller to decide if it's likely() or unlikely(), so make
the change to support that.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The XDP_REDIRECT packets are given to the XDP stack and
we drop the use of the related page: it will get freed
by the driver that ends up doing the Tx. Because we have
some hardware configurations with limited queue resources,
we use the existing datapath Tx queues rather than creating
and managing a separate set of xdp_tx queues.
Co-developed-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The XDP_TX packets get fed back into the Rx queue's partnered
Tx queue as an xdp_frame.
Co-developed-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set up the basics for running Rx packets through XDP programs.
Add new queue setup and teardown steps for adding/removing an
XDP program, and add the call to run the XDP on a packet.
The XDP frame size needs to be the MTU plus standard ethernet
header, plus head room for XDP scribblings and tail room for a
struct skb_shared_info. Also, at this point, we don't support
XDP frags, only a single contiguous Rx buffer. This means
that our page splitting is not very useful, so when XDP is in
use we need to use the full Rx buffer size and not do sharing.
Co-developed-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the checks are:
[1] unlikely(q->features & IONIC_TXQ_F_HWSTAMP)
[2] !unlikely(q->features & IONIC_TXQ_F_HWSTAMP)
[1] is clear enough, but [2] isn't exactly obvious to the
reader because !unlikely reads as likely. However, that's
not what this means.
[2] means that it's unlikely that the q has
IONIC_TXQ_F_HWSTAMP enabled.
Write an inline helper function to hide the unlikely
optimization to make these checks more readable.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Link: https://lore.kernel.org/r/20231204210936.16587-5-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add support for the PCI reset handlers in order to manage an FLR event.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Following the example of 'commit 9a0f830f80 ("ethtool: linkstate:
add a statistic for PHY down events")', added support for link down
events.
Add callback ionic_get_link_ext_stats to ionic_ethtool.c to support
link_down_count, a property of netdev that gets reported exclusively
on physical link down events.
Run ethtool -I <devname> to display the device link down count.
Signed-off-by: Nitya Sunkad <nitya.sunkad@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ionic device has on-board memory (CMB) that can be used
for descriptors as a way to speed descriptor access for faster
packet processing. It is rumored to improve latency and/or
packets-per-second for some profiles of small packet traffic,
although your mileage may vary.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In one version of the HW there is a remote possibility that it
will miss the doorbell ring. This adds a bit of protection to
be sure we don't stall a queue from a missed doorbell.
Fixes: 0f3154e6bc ("ionic: Add Tx and Rx handling")
Signed-off-by: Allen Hubbe <allen.hubbe@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The dbid_inuse bitmap is not useful in this driver so remove it.
Fixes: 6461b446f2 ("ionic: Add interrupts and doorbells")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Between fw running and fw actually stopped into reset, we need
a fw_stopping concept to catch and block some actions while
we're transitioning to FW_RESET state. This will help to be
sure the fw_up task is not scheduled until after the fw_down
task has completed.
On some rare occasion timing, it is possible for the fw_up task
to try to run before the fw_down task, then not get run after
the fw_down task has run, leaving the device in a down state.
This is possible if the watchdog goes off in between finding the
down transition and starting the fw_down task, where the later
watchdog sees the FW is back up and schedules a fw_up task.
Fixes: c672412f61 ("ionic: remove lifs on fw reset")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add vlans to the existing rx_filter_sync mechanics currently
used for managing mac filters.
Older versions of our firmware had no enforced limits on the
number of vlans that the LIF could request, but requesting large
numbers of vlans caused issues in FW memory management, so an
arbitrary limit was added in the FW. The FW now returns -ENOSPC
when it hits that limit, which the driver needs to handle.
Unfortunately, the FW doesn't advertise the vlan id limit,
as it does with mac filters, so the driver won't know the
limit until it bumps into it. We'll grab the current vlan id
count and use that as the limit from there on and thus prevent
getting any more -ENOSPC errors.
Just as is done for the mac filters, the device puts the device
into promiscuous mode when -ENOSPC is seen for vlan ids, and
the driver will track the vlans that aren't synced to the FW.
When vlans are removed, the driver will retry the un-synced
vlans. If all outstanding vlans are synced, the promiscuous
mode will be disabled.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The overflow flags really aren't useful and we don't need lif
struct elements to track them.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
These debug stats are not really useful, their collection is
likely detrimental to performance, and they suck up a lot
of memory which never gets used if no one ever enables the
priv-flag to print them, so just remove these bits.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The queues can be freed in ionic_close(). They need to be recreated
after ionic_open(). It doesn't need to replay the whole config. It
only needs to create the timestamping queues again.
Signed-off-by: Allen Hubbe <allenbh@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to separate the atomic needs of __dev_uc_sync()
and __dev_mc_sync() from the safe rx_mode handling, we need
to have the ndo handler manipulate the driver's filter list,
and later have the driver sync the filters to the firmware,
outside of the atomic context.
Here we put __dev_mc_sync() and __dev_uc_sync() back into the
ndo callback to give them their netif_addr_lock context and
have them update the driver's filter list, flagging changes
that should be made to the device filter list. Later, in the
rx_mode handler, we read those hints and sync up the device's
list as needed.
It is possible for multiple add/delete requests to come from
the stack before the rx_mode task processes the list, but the
handling of the sync status flag should keep everything sorted
correctly. For example, if a delete of an existing filter is
followed by another add before the rx_mode task is run, as can
happen when going in and out of a bond, the add will cancel
the delete and no actual changes will be sent to the device.
We also add a check in the watchdog to see if there are any
stray unsync'd filters, possibly left over from a filter
overflow and waiting to get sync'd after some other filter
gets removed to make room.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
With the move of mac filter handling to outside of the
ndo_rx_mode context using the IONIC_DW_TYPE_RX_MODE,
we no longer are using IONIC_DW_TYPE_RX_ADDR_ADD and
IONIC_DW_TYPE_RX_ADDR_DEL and they can be removed.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
If PTP configuration is attempted on ports that don't support
it, such as VF ports, the driver will return an error status
-95, or EOPNOSUPP and print an error message
enp98s0: hwstamp set failed: -95
Because some daemons can retry every few seconds, this can end
up filling the dmesg log and pushing out other more useful
messages.
We can catch this issue earlier in our handling and return
the error without a log message.
Fixes: 829600ce5e ("ionic: add ts_config replay")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the bulk of the code from ionic_set_rx_mode(), which
can be called from atomic context, into ionic_lif_rx_mode()
which is a safe context.
A call from the stack will get pushed off into a work thread,
but it is also possible to simultaneously have a call driven
by a queue reconfig request from an ethtool command or fw
recovery event. We add a mutex around the rx_mode work to be
sure they don't collide.
Fixes: 81dbc24147 ("ionic: change set_rx_mode from_ndo to can_sleep")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Split the call into ionic_lif_hwstamp_set() to have two
separate interfaces, one from the ioctl() for changing the
configuration and one for replaying the current configuration
after a FW RESET.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
These are changes to compile and link the new code, but no
new feature support is available or advertised yet.
Signed-off-by: Allen Hubbe <allenbh@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparating for hardware timestamping, we need to support
large Tx and Rx completion descriptors. Here we add the new
queue feature ids and handling for the completion descriptor
sizes.
We only are adding support for the Rx 2x sized completion
descriptors in the general Rx queues for now as we will be
using it for PTP Rx support, and we don't have an immediate
use for the large descriptors in the general Tx queues yet;
it will be used in a special Tx queues added in one of the
next few patches.
Signed-off-by: Allen Hubbe <allenbh@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up to now we've been ignoring any error return from the
queue starting in the link status check, so we fix that here.
If the driver had to reset and couldn't get things running
properly again, for example after a Tx Timeout and the FW is
not responding to commands, don't let the link watchdog try
to restart the queues. At this point the user can try to DOWN
and UP the device to clear the errors.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clean up a couple of struct uses to make for better fast path
access.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the 'ionic_reset_cb' typedef as it is not used.
Signed-off-by: Chen Lin <chen.lin5@zte.com.cn>
Acked-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
With a few more uses of true and false in function calls, we
need to give them some useful names so we can tell from the
calling point what we're doing.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The in_interrupt() usage in this driver tries to figure out which context
may sleep and which context may not sleep. in_interrupt() is not really
suitable as it misses both preemption disabled and interrupt disabled
invocations from task context.
Conditionals like that in driver code are frowned upon in general because
invocations of functions from invalid contexts might not be detected
as the conditional papers over it.
ionic_lif_addr() and _ionoc_lif_rx_mode() can be called from:
1) ->ndo_set_rx_mode() which is under netif_addr_lock_bh()) so it must not
sleep.
2) Init and setup functions which are in fully preemptible task context.
ionic_link_status_check_request() has two call paths:
1) NAPI which obviously cannot sleep
2) Setup which is again fully preemptible task context
Add arguments which convey the execution context to the affected functions
and let the callers provide the context instead of letting the functions
deduce it.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the dim library to manage dynamic interrupt
moderation in ionic.
v3: rebase
v2: untangled declarations in ionic_dim_work()
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move a few active struct fields to the front of the struct
for a little better cache use and performance.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert tx_timeout handler to not do the full reset. As this was
the last user of ionic_reset_queues(), we can drop it.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original way of changing ring length was to completely
tear down the lif's queue structure and then rebuild it, while
running the risk of allocations that might fail in the middle
and leave us with a broken driver.
Instead, we can set up all the new queue and descriptor
allocations first, then swap them out and delete the old
allocations. If the new allocations fail, we report the error,
stay with the old setup and continue running. This gives us
a safer path, and a smaller window of time where we're not
processing traffic.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use index counters rather than pointers for tracking head
and tail in the queues to save a little memory and to perhaps
slightly faster queue processing.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Split out the queue descriptor blocks into separate dma
allocations to make for smaller blocks.
Co-developed-by: Neel Patel <neel@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
ionic_open() and ionic_stop() are not referenced outside of their
defining file, so make them static.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use a block of stats structs attached to the lif instead of
little ones attached to each qcq. This simplifies our memory
management and gets rid of a lot of unnecessary indirection.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
As we aren't yet supporting multiple lifs, we can remove
complexity by removing the list concept and related code,
to be re-engineered later when actually needed.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the capability to split the Tx queues onto their own
interrupts with their own napi contexts. This gives the
opportunity for more direct control of Tx interrupt
handling, such as CPU affinity and interrupt coalescing,
useful for some traffic loads.
v2: use ethtool -L, not a vendor specific priv-flag
v3: simplify logging, drop unnecessary "no-change" tests
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
We give the tx clean path its own budget and service routine in
order to give a little more leeway to be more aggressive, and
in preparation for coming changes. We've found this gives us
a little better performance in some packet processing scenarios
without hurting other scenarios.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The UDP reuseport conflict was a little bit tricky.
The net-next code, via bpf-next, extracted the reuseport handling
into a helper so that the BPF sk lookup code could invoke it.
At the same time, the logic for reuseport handling of unconnected
sockets changed via commit efc6b6f6c3
which changed the logic to carry on the reuseport result into the
rest of the lookup loop if we do not return immediately.
This requires moving the reuseport_has_conns() logic into the callers.
While we are here, get rid of inline directives as they do not belong
in foo.c files.
The other changes were cases of more straightforward overlapping
modifications.
Signed-off-by: David S. Miller <davem@davemloft.net>
Change from using hardcoded MTU limits and instead use the
firmware defined limits. The value from the LIF attributes is
the frame size, so we take off the header size to convert to
MTU size.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ionic_wait_on_bit_lock() was a open-coded mutex knock-off
used only for protecting the queue reset operations, and there
was no reason not to use the real thing. We can use the lock
more correctly and to better protect the queue stop and start
operations from cross threading. We can also remove a useless
and expensive bit operation from the Rx path.
This fixes a case found where the link_status_check from a link
flap could run into an MTU change and cause a crash.
Fixes: beead698b1 ("ionic: Add the basic NDO callbacks for netdev support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
The queue reset pattern is used in a couple different places,
only slightly different from each other, and could cause
issues if one gets changed and the other didn't. This puts
them together so that only one version is needed, yet each
can have slighty different effects by passing in a pointer
to a work function to do whatever configuration twiddling is
needed in the middle of the reset.
This specifically addresses issues seen where under loops
of changing ring size or queue count parameters we could
occasionally bump into the netdev watchdog.
v2: added more commit message commentary
Fixes: 4d03e00a21 ("ionic: Add initial ethtool support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
