The TOP is a shared (between BT and WiFi) hardware component,
and if it has an error we need to reset the whole device, i.e.
both BT and WiFi. This is achieved by calling a specific ACPI
DSM (device-specific method) with the right arguments before
doing a reset via the object referenced by _PRR.
Since this is needed here, but a function reset will always do
better than just re-enumerating the bus in case of errors, we
can always try to at least do a function reset and do the full
product reset only when needed for TOP errors.
Also, for some Bz and Sc devices where BT is PCIe/IOSF as well,
find the BT device and unbind that device as well so the BT
driver can recover from the reset that's going to happen,
rather than having to somehow detect that the device was reset.
Also add - currently unused - the function reset mode, this is
going to get used in the upcoming escalation model.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20241231135726.5b0f846d3e13.Ia14ccac38ac3d48adf5f341b17c7e34ccc41c065@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
In order to later add the ability to do deeper resets of the
device when it crashes, first restructure the firmware error
handling. Instead of having just a single nic_error() method
that handles all, split it:
- nic_error() just handles and prints the error itself,
- dump_error() synchronously creates an error dump, and
- sw_reset() will be called to request doing a SW reset.
This changes the architecture so that the transport is now
responsible for deciding how to do the reset, and therefore
the handling of reprobe if error occurs during reconfig
moves there, which necessitates adding a method there that
notifies the transport that the recovery was completed.
Actually introducing the model under which deeper resets can
be done will be in future patches.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20241227095718.6d4f741ae907.I96a9243e7877808ed6d1bff6967c15d6c24882f0@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Instead of differentiating only sync/async, differentiate
the type of error, and document that only reset handshake
timeout (IWL_ERR_TYPE_RESET_HS_TIMEOUT) needs sync handling.
The special sync handling is somewhat temporary, the idea
is to later split the nic_error() method into error dump,
synchronizing the dump, and SW reset methods, and the type
is mostly in order to unify command queue full handling
into that new architecture as well.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20241227095718.aed9c9e4fac0.I2288042bec4728a75b61cb7f6ded5214bfa3ce85@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When the firmware crashes, we first told the op_mode and only then,
changed the transport's state. This is a problem if the op_mode's
nic_error() handler needs to send a host command: it'll see that the
transport's state still reflects that the firmware is alive.
Today, this has no consequences since we set the STATUS_FW_ERROR bit and
that will prevent sending host commands. iwl_fw_dbg_stop_restart_recording
looks at this bit to know not to send a host command for example.
To fix the hibernation, we needed to reset the firmware without having
an error and checking STATUS_FW_ERROR to see whether the firmware is
alive will no longer hold, so this change is necessary as well.
Change the flow a bit.
Change trans->state before calling the op_mode's nic_error() method and
check trans->state instead of STATUS_FW_ERROR. This will keep the
current behavior of iwl_fw_dbg_stop_restart_recording upon firmware
error, and it'll allow us to call iwl_fw_dbg_stop_restart_recording
safely even if STATUS_FW_ERROR is clear, but yet, the firmware is not
alive.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20240825191257.9d7427fbdfd7.Ia056ca57029a382c921d6f7b6a6b28fc480f2f22@changeid
[I missed this was a dependency for the hibernation fix, changed
the commit message a bit accordingly]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The TX queue code was mostly moved out to support an internal
transport that we were never going to publish, but we're no
longer using that. Since we're also going to be dissolving
the virtual transport layer entirely, integrate the TX queue
code into the PCIe layer.
This also has a small kernel of already removing the virtual
transport function layer, since iwl_trans_send_cmd() calls
iwl_trans_pcie_send_hcmd() directly now, even if that still
calls the transport send_cmd method for now, we'll clean it
up later.
Also, not everything is renamed yet.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://msgid.link/20240605140327.936b13f45071.Ib219ce01a1e67bcad79d5131626db950252aaa46@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
There's only one user of this code, which is STA unblock
during sleep for uAPSD on really old devices. Instead of
having this all through the API with calls up and down,
just implemented a special-case CMD_BLOCK_TXQS flag for
this, it's only needed in the old gen1 transport.
While at it, fix a complain that lockdep would have, as
we lock the cmd queue and then the TXQs in the reclaim
by using spin_lock_nested(). We no longer need to disable
BHs in iwl_trans_pcie_block_txq_ptrs() since it's called
with them disabled already.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://msgid.link/20231207044813.2bd95e0570fc.I16486dbc82570d2f73a585872f5394698627310d@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
If a TX queue has no space for new TX frames, the driver will keep
these frames in the overflow queue, and during reclaim flow it
will retry to send the frames from that queue.
But if the reclaim flow was invoked from TX queue flush, we will also
TX these frames, which is wrong as we don't want to TX anything
after flush.
This might also cause assert 0x125F when removing the queue,
saying that the driver removes a non-empty queue
Fix this by TXing the overflow queue's frames only if we are
not in flush queue flow.
Fixes: a445098058 ("iwlwifi: move reclaim flows to the queue file")
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20231022173519.caf06c8709d9.Ibf664ccb3f952e836f8fa461ea58fc08e5c46e88@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Multi rx queue allows to spread the load of the Rx streams on different
CPUs. 9000 series required complex synchronization mechanisms from the
driver side since the hardware / firmware is not able to provide
information about duplicate packets and timeouts inside the reordering
buffer.
Users have complained that for newer devices, all those synchronization
mechanisms have caused spurious packet drops. Those packet drops
disappeared if we simplify the code, but unfortunately, we can't have
RSS enabled on 9000 series without this complex code.
Remove support for RSS on 9000 so that we can make the code much simpler
for newer devices and fix the bugs for them.
The down side of this patch is a that all the Rx path will be routed to
a single CPU, but this has never been an issue, the modern CPUs are just
fast enough to cope with all the traffic.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20231017115047.2917eb8b7af9.Iddd7dcf335387ba46fcbbb6067ef4ff9cd3755a7@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Generalize iwl_pnvm_parse(). This saves us from copying each payload
twice (first in the parsing and later when copying it to the dram).
Moreover, its more compatible for handling larger pnvm tables in
the future (in which payloads won't be concatenated).
The main changes are:
1. Take out the concatenating of the payloads from the parsing level
2. Start using iwl_pnvm_image structure that will hold pointers to
payloads that should be delivered to fw, their sizes and number.
Signed-off-by: Alon Giladi <alon.giladi@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20230606103519.06c02f380b6f.I03a3030fca194aa0c4bc2ecd18531f8914e98cfd@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Newer firmware versions will support a new queue allocation
command, in order to deal with MLD where multiple stations
are used for a single queue. Add support for the new command.
This requires some refactoring of the queue allocation API,
which now gets
- the station mask instead of the station ID
- the flags without the "enable" flag, since that's no longer
used in the new API
Additionally, this new API now requires that we remove queues
before removing a station, the firmware will no longer do that
internally. Also add support for that.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20220210181930.acbf22ac2b66.I2bf38578c5ca1f7ffb2011a782f772db92fc4965@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
