Instead of calculating the offset to the 802.11 header based on radiotap
bits and length, shorten the code path by always setting the MAC header
in the skb and using skb_mac_header().
Signed-off-by: Mordechay Goodstein <mordechay.goodstein@intel.com>
Link: https://lore.kernel.org/r/20230127002430.3ec5493934a4.I1d41a2af28588b5899fcd2402f8c4bd8cc29a12e@changeid
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Now with all the prework, this is fairly simple, just report the
new bandwidth and RX_ENC_EHT type in RX, and for now just do a
minimal report of the EHT TLC rate in iwl_mvm_set_sta_rate().
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20230109130329.5f34d73d1f74.Ib27ae7bd23bc152d61021fd73aabdc76679b9fe4@changeid
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
In some rare occasions, the firmware may let some frames with invalid
rates, such as CCK rates on the high band, come through. This causes
the driver to issue a warning, but since this is a possible issue and
it's not really a bug in the driver, convert the warning into an
error.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20221102165239.3d3673c70556.I13464b11d405fd6021618b0a32404cecb7e9ac51@changeid
There's a bunch of duplicated code in the normal data RX and
the no-data RX paths, refactor that.
Note that we're evidently not implementing the 6 GHz band in
the no-data path correctly, need to fix that in firmware
first. Also, we were setting the NSS differently, keep doing
that, but just override the previous values.
Note also that we used to drop frames with bad rate data, to
simplify that just report rate 0 and continue.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20220906161827.a81580d5e769.I3ee7a1fe1adf8684c48713dbbdc2cbc60bd24cd1@changeid
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
There's currently only one driver that reports CHECKSUM_COMPLETE,
that is iwlwifi. The current hardware there calculates checksum
after the SNAP header, but only RFC 1042 (and some other cases,
but replicating the exact hardware logic for corner cases in the
driver seemed awkward.)
Newer generations of hardware will checksum _including_ the SNAP,
which makes things easier.
To handle that, simply always assume the checksum _includes_ the
SNAP header, which this patch does, requiring to first add it
for older iwlwifi hardware, and then remove it again later on
conversion.
Alternatively, we could have:
1) Always assumed the checksum starts _after_ the SNAP header;
the problem with this is that we'd have to replace the exact
"what is the SNAP" check in iwlwifi that cfg80211 has.
2) Made it configurable with some flag, but that seemed like too
much complexity.
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.20220202104617.230736e19e0e.I3e6745873585ad943c152fab9e23b5221f17a95f@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The RADA might include in the Rx frame the MIC and CRC bytes.
These bytes should be removed for non monitor interfaces and
should not be passed to mac80211.
Fix the Rx processing to remove the extra bytes on non monitor
cases.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211219121514.098be12c801e.I1d81733d8a75b84c3b20eb6e0d14ab3405ca6a86@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
On Bz devices, the hardware checksums including the SNAP header,
starting directly after the MAC header, so we don't need the
extra checks and can just pass the checksum to mac80211.
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.20211219121514.8ef59da4e05e.Ide87c2dc8fa08eae55c013a625f4ece5184b1b63@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
This line was indented too far and was a bit confusing.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/20211110080759.GE5176@kili
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
If we happen to decide an NSSN queue sync (IWL_MVM_RXQ_NSSN_SYNC)
for some remaining packets that are still on the queue, but just
after we've decided to do a delBA (which causes its own queues
sync with IWL_MVM_RXQ_NOTIF_DEL_BA) we can end up with a sequence
of events like this:
CPU 1 CPU 2
remove BA session with baid N
send IWL_MVM_RXQ_NOTIF_DEL_BA
send IWL_MVM_RXQ_NSSN_SYNC
get IWL_MVM_RXQ_NOTIF_DEL_BA
get IWL_MVM_RXQ_NOTIF_DEL_BA
get IWL_MVM_RXQ_NSSN_SYNC
complete IWL_MVM_RXQ_NOTIF_DEL_BA
remove N from baid_map[]
get IWL_MVM_RXQ_NSSN_SYNC
WARN_ON(!baid_map[N])
Thus, there's a race that leads in hitting the WARN_ON, but more
importantly, it's a race that potentially even results in a new
aggregation session getting assigned to baid N.
To fix this, remove the WARN_ON() in the NSSN_SYNC case, we can't
completely protect against hitting this case, so we shouldn't be
warning. However, guard ourselves against BAID reuse by doing yet
another round of queue synchronization after the entry is removed
from the baid_map, so that it cannot be reused with any in-flight
IWL_MVM_RXQ_NSSN_SYNC messages.
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.20211204083237.44abbbc50f40.I5492600dfe513356555abe2d7df0e2835846e3d8@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Remove the unused csi parameter from
iwl_mvm_pass_packet_to_mac80211(). It is not used anymore, but was
accidentally left in.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/20211017132604.480251-2-luca@coelho.fi
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
As part of the new rate_n_flags this two APIs
may use the new rate. Add support for the new rate.
These two APIs were updated in one patch because both of them
are using the same functions which were changed.
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211017162352.7eb6240b8079.I1e804e70a8ebd23840a9431fc5d2a56ad8d5d1a2@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
To check beacon protection had a mismatch of the key ID we
currently use the key ID from the firmware, but firmware
side we want to clean up the API to stop reporting this.
Instead, check the IWL_RX_MPDU_STATUS_KEY_VALID bit that
indicates that the firmware used the correct key, and if
that's set but we get invalid MIC/replay use the key ID
from the frame to look up the key and notify mac80211 of
MIC error or replay. Since both keys must have the same
cipher and thus MIC length, we can use either of them to
look up the MIC length.
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.20210826224715.30e665d39b07.I78bf7d304ef5a80cecf5fa1c1fca0b51b956cceb@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
When a station is removed, the driver-mac80211 station mapping is removed
before the station is actually deleted from the FW. As a result, it is
reasonable that the FW will continue to pass frames although the driver
doesn't have a station for them anymore. Thus change the message
severity level from ERR to DEBUG_DROP.
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20210805130823.a751177743c6.I5607504dade88ba461508643f58390dd661c05ba@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
We don't fill in phy_data.d1 in no-data RX, and thus we
pretend some data is actually filled in radiotap when it
isn't or has default (zero) values.
Fill in phy_data.d1 appropriately, and while at it also
move the info_type initialization into the initializer.
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.20210618105614.3d488885f77c.Ib97a2bc57c1e9fb98927dc6f802568db313abe3b@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
When we do queue sync, it's confusing that we have the structures
declared in the FW API header files that aren't really firmware,
and the union is also confusing - especially now in the code that
checks the size on the return.
So rework this: change the type of sync and whether to do it in a
synchronous fashion to arguments, and build the data structure in
the function, so we don't need the union.
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.20210331121101.f62833fd9893.I612d7ac1c655ec4880329360e15d207698c750bc@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Unfortunately, since beacon protection isn't fully available
yet, we didn't notice that there are problems with it and
that the replay detection isn't working correctly. We were
relying only on mac80211, since iwl_mvm_rx_crypto() exits
when !ieee80211_has_protected(), which is of course true for
protected (but not encrypted) management frames.
Fix this to properly detect protected (but not encrypted)
management frames and handle them - we continue to only care
about beacons since for others everything can and will be
checked in mac80211.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Fixes: b1fdc2505a ("iwlwifi: mvm: advertise BIGTK client support if available")
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/iwlwifi.20210326125611.23c990843369.I09c262a8f6f9852cc8f513cdcb31a7f8f87dd8af@changeid
Once the all the stations completed the switch, we need
to clear csa_tx_blocked_vif. This was missing. We also
need to re-enable the broadcast / multicast stations.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20210205110447.f5b813753bdb.Id58979b678974c3ccf44d8b381c68165ac55a3d3@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
If the firmware has support, then advertise it to the stack and
send the key down. Since we re-check the protection in the host
anyway, we don't really need to do anything on RX except that we
should drop frames that the firmware _knows_ are replay errors,
since beacon filtering might otherwise result in replays being
possible.
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.20210205110447.f5a3d53301b3.I23e84c9bb0b039d9106a07e9d6847776757f9029@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
In preparation to enable -Wimplicit-fallthrough for Clang, fix multiple
warnings by explicitly using the fallthrough pseudo-keyword as a
replacement for a number of "fall through" markings.
Notice that Clang doesn't recognize "fall through" comments as
implicit fall-through.
Link: https://github.com/KSPP/linux/issues/115
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20201117135053.GA13248@embeddedor
Previously, we added checks that the contained MPDU size is long
enough, but really we should also check that the notification
itself fits into the data. Add some checks for that.
Also add unlikely() annotations on the previously added checks.
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.20201209231352.51cc04cf1e3e.I7bfd6809f8f5feb75f79397646e6656e95688a0e@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
We send some data to the firmware and expect to get it back,
but we shouldn't really trust the firmware on this. Check the
size of all the data we send down to avoid using bad or just
uninitialized data when the firmware doesn't respond right.
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.20201209231352.a5a8173f16c7.I4fa68bb2b1c7dcc52ddd381c4042722d27c4a34d@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
We're currently doing accounting on the queue sync with an
atomic variable that counts down the number of remaining
notifications that we still need.
As we've been hitting issues in this area, modify this to
track a bitmap of queues, not just the number of queues,
and print out the remaining bitmap in the warning.
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.20201209231352.0a3fa177cd6b.I7c69ff999419368266279ec27dd618eb450908b3@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
The argument type to iwl_mvm_notify_rx_queue() is currently just
a u8 *, but that's misleading because we actually need the inner
data to be of type struct iwl_mvm_internal_rxq_notif, because we
interpret it when we get it back from the device (to check the
sync bool and possibly the cookie.)
Therefore, clear up any potential confusion and require that the
data passed is of type struct iwl_mvm_internal_rxq_notif *.
Also, while at it, rename the "count" to "notif_size" as "count"
doesn't really clearly say what it's counting.
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.20201209231352.d28e14682bdc.I9ac366aa97db045be4daa4ba263267a3ac6a6a2f@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
The MPDU contained in a notification shouldn't be larger than the
notification size itself is, validate this.
Reported-by: Haggai Abramovsky <haggai.abramovsky@intel.com>
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.20201209231352.7c721ad37014.Id5746874ecfa208b60baa62691b2d9dc5dd4d89c@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Add support to the 6GHz band (aka. Ultra High Band or UHB). This
allows us to scan and connect to channels in that band, including all
the relevant features, such as preferred scan channels, colocated
channels etc.
Co-developed-by: Haim Dreyfuss <haim.dreyfuss@intel.com>
Signed-off-by: Haim Dreyfuss <haim.dreyfuss@intel.com>
Co-developed-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Co-developed-by: Tova Mussai <tova.mussai@intel.com>
Signed-off-by: Tova Mussai <tova.mussai@intel.com>
Co-developed-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Co-developed-by: Tali Levi Rovinsky <Tali.Levi-rovinsky@intel.com>
Signed-off-by: Tali Levi Rovinsky <Tali.Levi-rovinsky@intel.com>
Co-developed-by: Avraham Stern <avraham.stern@intel.com>
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Co-developed-by: Ayala Beker <ayala.beker@intel.com>
Signed-off-by: Ayala Beker <ayala.beker@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20201210000657.0fdbfc3d7352.Idb648536faf21716e2ab2c6d6890d3e49f719cd3@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
There are some bits declared here that simply don't exist
in the firmware, and some are missing (e.g. the key) from
what the firmware has. Align this and move all the fields
into a single one for this status word, which makes this a
bit easier.
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.20200930191738.0ba403d72e7c.I5fa3aa0538f3fbf8c3885b27a1204b5b0464c20a@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
On newer hardware, we have the full checksum, so use it to report
CHECKSUM_COMPLETE and avoid the protocol specific hardware parsing.
Note that the hardware already parses/removes the SNAP header, so
we actually literally get what we need to report to the stack, as
we're expected to checksum everything after the L2 header (which
is translated/added by mac80211).
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.20200926002540.869e829c815d.I70f374865b0acafc675a8d7959912eeaeb595acf@changeid
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
The purpose of this was to keep all the queues updated with
the Rx sequence numbers because unlikely yet possible
situations where queues can't understand if a specific
packet needs to be dropped or not.
Unfortunately, it was reported that this caused issues in
our DMA engine. We don't fully understand how this is related,
but this is being currently debugged. For now, just don't send
this notification to the Rx queues. This de-facto reverts my
commit 3c514bf831:
iwlwifi: mvm: add a loose synchronization of the NSSN across Rx queues
This issue was reported here:
https://bugzilla.kernel.org/show_bug.cgi?id=204873https://bugzilla.kernel.org/show_bug.cgi?id=205001
and others maybe.
Fixes: 3c514bf831 ("iwlwifi: mvm: add a loose synchronization of the NSSN across Rx queues")
CC: <stable@vger.kernel.org> # 5.3+
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
This email alias (ilw@linux.intel.com) hasn't been functional
for probably closer to a decade than not, remove it. It's not
really clear to me how this ended up in new code though.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
This is dead code, nothing uses the IWL_DEVICE_22560 macro and
thus nothing every uses IWL_DEVICE_FAMILY_22560. Remove it all.
While at it, remove some code and definitions used only in this
case, and clean up some comments/names that still refer to it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The FW introduce new API to get the band from the rx mpdu,
use this new API.
Signed-off-by: Tova Mussai <tova.mussai@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Sometimes the firmware won't be able to decrypt frames
because the keys were not installed yet or other scenarios.
The firmware will soon stop dropping multicast frames when
MAC_FILTER_ACCEPT_GRP is not set. The firmware will simply
always pass multicast frame in.
In order to avoid logging any such frame coming in when we
don't have the keys, drop the print.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
In prior hardware generations (e.g. 9000 series), we received the BAR
frame with fake NSSN information to handle releasing frames from the
reorder buffer for the default queue, the other queues were getting
the FRAME_RELEASE notification in this case.
With multi-TID block-ack, the firmware no longer sends us the BAR
frame because the fake RX is quite big (just the metadata is around
48 bytes or so). Instead, it now sends us one (or multiple) special
release notifications (0xc2). The hardware consumes these as well,
but only generates the FRAME_RELEASE (0xc3) for queues other than
the default queue. We thus need to handle them in the same way we
handle the normal FRAME_RELEASE.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Certain APs (I think a certain Broadcom model) interact badly with our
full state BA bitmap handling, and if triggered badly with many powersave
transitions they keep sending frames from before the window, which our
hardware then doesn't appear to ACK (to them) since it has moved on and
is sending ACKs for higher SNs now.
Try to detect this situation and if this keeps happening, disable the
aggregation session.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Add a pointer to the iwl_trans structure and point it to the trans
part of the cfg. This is the first step in disassociating the trans
configuration from the rest of the configuration.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
In order to be able to select the cfg depending on the HW revision or
on the RF ID, we need to set up the trans before selecting the cfg.
To do so, move the elements from cfg that are needed by
iwl_trans_alloc() to a separate struct at the top of the cfg, so it
can be used by other cfg types as well, before selecting the rest of
the configuration.
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
An earlier patch made sure that the queues are not lagging
too far behind. This means that iwl_mvm_release_frames
should not be called with a head_sn too far behind NSSN.
Don't take the risk to change completely the entry
condition to iwl_mvm_release_frames, but don't update
the head_sn is the NSSN is more than 2048 packets ahead
of us. Since this just cannot be right. This means that
the scenario described here happened. We are queue 0.
Q:0 Q:1
head_sn: 0 -> 2047
head_sn: 2048
Lots of packets arrive:
head_sn: 2047 -> 2150
send NSSN_SYNC notification
Handle notification
from the firmware and
do NOT move the head_sn
back to 2048
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
In order to support MSI-X efficiently, we want to avoid
communication across Rx queues. Each Rx queue should have
all the data it needs to process a packet.
The reordering buffer is a challenge in the MSI-X world
since we can have a single BA session whose packets are
directed to different queues. This is why each queue has
its own reordering buffer. The hardware is able to hint
the driver whether we have a hole or not, which allows
the driver to know whether it can release a packet or not.
This indication is called NSSN. Roughly, if the packet's
SN is lower than the NSSN, we can release the packet to
the stack. The NSSN is the SN of the newest packet received
without any holes + 1.
This is working as long as we don't have packets that we
release because of a timeout. When that happens, we could
have taken the decision to release a packet after we have
been waiting for its predecessor for too long. If this
predecessor comes later, we have to drop it because we
can't release packets out of order. In that case, the
hardware will give us an indication that we can we release
the packet (SN < NSSN), but the packet still needs to be
dropped.
This is why we sometimes need to ignore the NSSN and we
track the head_sn in software.
Here is a specific example of this:
1) Rx queue 1 got packets: 480, 482, 483
2) We release 480 to to the stack and wait for 481
3) NSSN is now 481
4) The timeout expires
5) We release 482 and 483, NSSN is still 480
6) 481 arrives its NSSN is 484.
We need to drop 481 even if 481 < 484. This is why we'll
update the head_sn to 484 at step 2. The flow now is:
1) Rx queue 1 got packets: 480, 482, 483
2) We release 480 to to the stack and wait for 481
3) NSSN is now 481 / head_sn is 481
4) The timeout expires
5) We release 482 and 483, NSSN is still 480 but head_sn is 484.
6) 481 arrives its NSSN is 484, but head_sn is 484 and we drop it.
This code introduces another problem in case all the traffic
goes well (no hole, no timeout):
Rx queue 1: 0 -> 483 (head_sn = 484)
Rx queue 2: 501 -> 4095 (head_sn = 0)
Rx queue 2: 0 -> 480 (head_sn = 481)
Rx queue 1: 481 but head_sn = 484 and we drop it.
At this point, the SN of queue 1 is far behind: more than
4040 packets behind. Queue 1 will consider 481 "old"
because 481 is in [501-64:501] whereas it is a very new
packet.
In order to fix that, send an Rx notification from time to
time (twice across the full set of 4096 packets) to make
sure no Rx queue is lagging too far behind.
What will happen then is:
Rx queue 1: 0 -> 483 (head_sn = 484)
Rx queue 2: 501 -> 2047 (head_sn = 2048)
Rx queue 1: Sync nofication (head_sn = 2048)
Rx queue 2: 2048 -> 4095 (head_sn = 0)
Rx queue 1: Sync notification (head_sn = 0)
Rx queue 2: 1 -> 481 (head_sn = 482)
Rx queue 1: 481 and head_sn = 0.
In queue 1's data, head_sn is now 0, the packet coming in
is 481, it'll understand that the new packet is new and it
won't be dropped.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
