command len is set too early in the code, since when building
AMSDU, the size changes. This causes the byte count table to
have the wrong size.

Fixes: a0ec0169 ("iwlwifi: support new tx api")
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Currently code sets the write pointer when getting the TX queue
allocate response. This causes a redundant interrupt with any actual
change in the pointer. Remove this write altogether.

Fixes: 310181ec ("iwlwifi: move to TVQM mode")
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Split TX tracing to be per TB. This is needed now that
AMSDUs can be sent and skb can be larger than trace
limit.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When we TX AMSDU, we shouldn't pad the packet. In the past,
we were building AMSDU only in transport layer, and gen2
functions are built based on this. However, now that op mode
may build AMSDUs, we need to take care of padding also in
gen2 "non-pcie-amsdu" path.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

If we use the iwl_pcie_gen2_set_tb() return value for BIT(),
we should validate that it's not going to be negative, so do
the check and bail out if we hit an error. We shouldn't, as
we check if it'll fit beforehand, but better be safe.

Fixes: ab6c6445 ("iwlwifi: pcie: copy TX functions to new transport")
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

If the incoming frame should be an A-MSDU, it may already be one,
for example in the case of NAN multicast being encapsulated in an
A-MSDU. Thus, use the GSO algorithm to build A-MSDU only if the
skb actually contains GSO data.

Fixes: 6ffe5de3 ("iwlwifi: pcie: add AMSDU to gen2")
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Move the skb fragment loop into a helper routine to be able
to reuse it later.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Support more txq_alloc command types by moving the command declaration
to the gen specific area.  While at it, move some of the code segments
to a common place for re-use.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Allow other device generations to use the utilities that
are used to send and reclaim host commands and to allocate
rx, by making it non-static.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

We would like to allow using tx init code for other queues but
the command queue - for newer devices.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The code is different enough to justify a split.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

If CONFIG_IPV6 is not enabled in the kernel, tcp.h is not included
implicitly from other header files, causing compilation errors.  To
solve that, explicitly include it in tx-gen2.c.
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

22560 devices use a new tx cmd api. Update the code to use
the new api.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

22560 devices tfd queue max size is 2^16. Allow a configurable
max size in the driver for supporting different devices.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

For devices which use the image loader image, the length of the frame
must be updated in the byte count in bytes, and not dwords as today.
Avoid dividing the input length by 4.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Avoid a race where two (or more) commands get the
same index:

1. T1 calls enqueue_hcmd and the local TFD index is assigned to
   txq->write_ptr;
2. Context switch 'before incrementing txq->write_ptr';
3. T2 calls enqueue_hcmd and the local TFD index is assigned to
   txq->write_ptr;
4. Now the index is set to the same value for both commands of T1 and
   T2.

To prevent this from happening, set the local TFD index inside the
critical section (the index is set by global txq write pointer).
Signed-off-by: NShaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Op mode will begin tp use varying size of TX queue.
All the infra is in place, allow it.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

As preparation for dynamic queue sizing, add a parameter
of the TX queue size to the dynamic queue allocation op
mode API.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This reverts commit dd05f9aa.

Shorter TX queues support was added eventually without the
need for the parameters this patch added.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When support for shorter TX queues was introduced, it
didn't include the actual allocation of shorter queue,
which is the main motive for the change.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When we enable TSO, we can have a lot of packets in the
operation mode that will be pushed to the transport
no matter what is the queue's fullness state.

To cope with that the transport can buffer those packets
and add them to the ring later when there is more room.
This implementation was missing in the Gen2 devices'
code.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Our Transmit Frame Descriptor (TFD) is a DMA descriptor that
includes several pointers to be able to transmit a packet
which is not physically contiguous.

Depending on the hardware being use, we can have 20 or 25
pointers in a single TFD. In both cases, it is more than
enough and it is quite hard to hit this limit.
It has been reported that when using specific applications
(Ktorrent), we can actually use all the pointers and then
a long standing bug showed up.

When we free the TFD, we check its number of valid pointers
and make sure it doesn't exceed the number of pointers the
hardware support.
This check had an off by one bug: it is perfectly valid to
free the 20 pointers if the TFD has 20 pointers.

Fix that.

https://bugzilla.kernel.org/show_bug.cgi?id=197981Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

22000 devices (previously referenced as A000) can support
short transmit queues. This means that we have less DMA
descriptors (TFD) for those shorter queues.
Previous devices must still have 256 TFDs for each queue
even if those 256 TFDs point to fewer buffers.

When I introduced support for the short queues for 22000
I broke older devices by assuming that they can also have
less TFDs in their queues. This led to several problems:

1) the payload of the commands weren't unmapped properly
   which caused the SWIOTLB to complain at some point.
2) the hardware could get confused and we get hardware
   crashes.

The corresponding bugzilla entries are:

https://bugzilla.kernel.org/show_bug.cgi?id=198201
https://bugzilla.kernel.org/show_bug.cgi?id=198265

Cc: stable@vger.kernel.org # 4.14+
Fixes: 4ecab561 ("iwlwifi: pcie: support short Tx queues for A000 device family")
Reviewed-by: NSharon, Sara <sara.sharon@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NKalle Valo <kvalo@codeaurora.org>

This variable is never used, so remove the code to set it.
After this, the variable 'iph' also has the same fate.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Devices in the A000 family can use a different size for the command queue.
To allow this, make the command queue size configurable and set the size
for A000 devices to 32.
Signed-off-by: NShahar S Matityahu <shahar.s.matityahu@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This allows to modify TFD_TX_CMD_SLOTS to a power of 2
which is smaller than 256.
Note that we still need to set values to wrap at 256
into the scheduler's write pointer, but all the rest of
the code can use shorter transmit queues.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When we unmap a non-empty Tx queue, we need to free the
pages that we allocated for the headers in TSO flows.
This code existed for the 9000 device family, but somehow
it got left out when the new Tx path for the A000 devices
was written.

Fixes: 2b0c5946d9ed ("iwlwifi: pcie: introduce a000 TX queues management")
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The TVQM tells us the initial write pointer for a queue,
but that write pointer is in WiFi sequence number unit
and not in TFD index unit. Which means that the write
pointer in the TVQM's response can be bigger than the
Tx queue ring size.

Fix that by modulo'ing the write pointer from the TVQM
with the Tx queue size.

Fixes: 66128fa08806 ("iwlwifi: move to TVQM mode")
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The return status check of iwl_pcie_gen2_build_amsdu
was buggy. Fix it.

Fixes: 6ffe5de3 ("iwlwifi: pcie: add AMSDU to gen2")
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Apart from DVM, all firmware uses the same base API, and there's
code outside iwlmvm that needs to interact with it. Reflect this
in the source better and reorganize the firmware API to a new
fw/api/ directory.

While at it, split the already pretty large fw-api.h file into a
number of smaller files, going from almost 3k lines in there to
a maximum number of lines less than 1k.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When sending non-linear SKBs that should be included in the regular
TX tracing completely (and not be pushed into the tx_data tracing),
the (tracing) code didn't correctly take the fact that they were
non-linear into account and added only the skb head portion.

This probably never really triggered, since those frames we want
traced fully are most likely linear anyway, but the code gets easier
to understand and we lose an argument to the tracing function, so
overall fixing this is better.

Fixes: 206eea78 ("iwlwifi: pcie: support frag SKBs")
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

There's no need to calculate the data_len outside of the tracepoint,
since it's always skb->len - hdr_len, which are both available inside.
Simplify the callers and move the calculation in.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When toggling the RF-kill pin quickly in succession, the driver can
get rather confused because it might be in the process of shutting
down, expecting all commands to go through quickly due to rfkill,
but the transport already thinks the device is accessible again,
even though it previously shut it down. This leads to bugs, and I
even observed a kernel panic.

Avoid this by making the PCIe code only report that the radio is
enabled again after the higher layers actually decided to shut it
off.

This also pulls out this common RF-kill checking code into a common
function called by both transport generations and also moves it to
the direct method - in the internal helper we don't really care
about the RF-kill status anymore since we won't report it up until
the stop anyway.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The queue ID should never be 512 either, so correct the check
to be >= instead of just >.

Fixes: 310181ec ("iwlwifi: move to TVQM mode")
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This is essentially the same code as gen1, except that it uses
gen2 functions and SW checksum is not included.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Sending host command with CMD_WANT_SKB flag demands the release of the
response buffer with iwl_free_resp function.
The patch adds the memory release in all the relevant places
Signed-off-by: NShahar S Matityahu <shahar.s.matityahu@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This API replaces the complex NVM parsing of the iwlwifi module.
Instead, we get all needed data from firmware.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

In TVQM firmware returns the value of the queue ID and code
should accept it.
The TX queue config API was changed. Move to new API.
This has to be done in parallel in mvm and pcie.
Do not move yet to 512 queues since there are some opens
with enabling it.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

In TVQM mode the queue ID is assigned after enablement.
Get rid of assuming pre-defined TX queue ID in functions
that will be used by TVQM allocation path.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Change queue allocation to be dynamic. On transport init only
the command queue is being allocated. Other queues are allocated
on demand.
This is due to the huge amount of queues we will soon enable (512)
and as a preparation for TX Virtual Queue Manager feature (TVQM),
where firmware will assign the actual queue number on demand.
This includes also allocation of the byte count table per queue
and not as a contiguous chunk of memory.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>