Johannes Berg [Tue, 19 Jul 2022 10:08:29 +0000 (12:08 +0200)]
wifi: mac80211: sta_info: fix link_sta insertion
When inserting a link STA, make sure it doesn't exist first
and add lockdep assertions that we cannot modify the hash
table without holding the sta_mtx, so this check is really
correct.
Also return without hashing if the driver failed, and warn
if the hashing fails, which shouldn't happen due to the
check described above.
Fixes: cb71f1d136a6 ("wifi: mac80211: add sta link addition/removal") Fixes: ba6ddab94fc6 ("wifi: mac80211: maintain link-sta hash table") Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg [Tue, 19 Jul 2022 08:26:50 +0000 (10:26 +0200)]
wifi: nl80211/mac80211: clarify link ID in control port TX
Clarify the link ID behaviour in control port TX, we need it
to select the link to transmit on for both MLD and non-MLD
receivers, but select the link address as the SA only if the
receiver is not an MLD.
Fixes: 67207bab9341 ("wifi: cfg80211/mac80211: Support control port TX from specific link") Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg [Tue, 19 Jul 2022 07:37:59 +0000 (09:37 +0200)]
wifi: mac80211: return error from control port TX for drops
If the frame is going to be dropped anyway because
ieee80211_lookup_ra_sta() returned an error (and
even though it's a bit racy, it will likely continue
to do so), return the error out instead of just
silently dropping the frame.
Johannes Berg [Mon, 18 Jul 2022 21:01:19 +0000 (23:01 +0200)]
wifi: mac80211: more station handling sanity checks
Add more sanity checks to the API handling, we shouldn't
be able to create a station without links, nor should we
be able to add a link to a station that wasn't created as
an MLD with links in the first place.
Johannes Berg [Mon, 18 Jul 2022 20:54:27 +0000 (22:54 +0200)]
wifi: mac80211: fix link sta hash table handling
There are two issues here: we unhash the link stations only
directly before freeing the station they belong to, and we
also don't unhash all the links correctly in all cases. Fix
these issues.
Johannes Berg [Mon, 18 Jul 2022 14:04:05 +0000 (16:04 +0200)]
wifi: mac80211: mlme: set sta.mlo to mlo state
At this point, we've already changed link_id to be zero for
a non-MLO connection, so use the 'mlo' variable rather than
link ID to determine the MLO status of the station.
Fixes: bd363ee53302 ("wifi: mac80211: mlme: set sta.mlo correctly") Fixes: 81151ce462e5 ("wifi: mac80211: support MLO authentication/association with one link") Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg [Mon, 18 Jul 2022 14:58:18 +0000 (16:58 +0200)]
wifi: mac80211: fast-xmit: handle non-MLO clients
If there's a non-MLO client, the A2 must be set to the BSSID
of the link since no translation will happen in lower layers
and it's needed that way for encryption.
Johannes Berg [Mon, 18 Jul 2022 14:45:50 +0000 (16:45 +0200)]
wifi: mac80211_hwsim: fix address translation for MLO
There are two issues here: we need to do the translation
even in case mac80211 selected a link, and we should only
translate the A3 if it's the BSSID. Fix both.
Johannes Berg [Mon, 18 Jul 2022 07:14:30 +0000 (09:14 +0200)]
wifi: mac80211: fix NULL pointer deref with non-MLD STA
If we have a non-MLD STA on an AP MLD, we crash while
adding the station. Fix that, in this case we need to
use the STA's address also on the link data structure.
Fixes: f36fe0a2df03 ("wifi: mac80211: fix up link station creation/insertion") Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg [Sun, 17 Jul 2022 16:16:51 +0000 (18:16 +0200)]
wifi: mac80211: mlme: fix override calculation
In my previous changes here, I neglected to take the old
conn_flags into account that might still be present from
the authentication, and thus ieee80211_setup_assoc_link()
can misbehave, as well as the override calculation being
wrong. Fix that by ORing in the old flags.
Fixes: 1845c1d4a455 ("wifi: mac80211: mlme: refactor assoc link setup") Signed-off-by: Johannes Berg <[email protected]>
wifi: mac80211: do not abuse fq.lock in ieee80211_do_stop()
lockdep complains use of uninitialized spinlock at ieee80211_do_stop() [1],
for commit f856373e2f31ffd3 ("wifi: mac80211: do not wake queues on a vif
that is being stopped") guards clear_bit() using fq.lock even before
fq_init() from ieee80211_txq_setup_flows() initializes this spinlock.
According to discussion [2], Toke was not happy with expanding usage of
fq.lock. Since __ieee80211_wake_txqs() is called under RCU read lock, we
can instead use synchronize_rcu() for flushing ieee80211_wake_txqs().
The assignment of the value to the variable total in the loop
condition must be enclosed in additional parentheses, since otherwise,
in accordance with the precedence of the operators, the conjunction
will be performed first, and only then the assignment.
Due to this error, a warning later in the function after the loop may
not occur in the situation when it should.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
When building with Clang we encounter this warning:
| drivers/net/wireless/mediatek/mt7601u/debugfs.c:92:6: error: format
| specifies type 'unsigned char' but the argument has type 'int'
| [-Werror,-Wformat] dev->ee->reg.start + dev->ee->reg.num - 1);
The format specifier used is `%hhu` which describes a u8. Both
`dev->ee->reg.start` and `.num` are u8 as well. However, the expression
as a whole is promoted to an int as you cannot get smaller-than-int from
addition. Therefore, to fix the warning, use the promoted-to-type's
format specifier -- in this case `%d`.
example:
```
uint8_t a = 4, b = 7;
int size = sizeof(a + b - 1);
printf("%d\n", size);
// output: 4
```
See more:
(https://wiki.sei.cmu.edu/confluence/display/c/INT02-C.+Understand+integer+conversion+rules)
"Integer types smaller than int are promoted when an operation is
performed on them. If all values of the original type can be represented
as an int, the value of the smaller type is converted to an int;
otherwise, it is converted to an unsigned int."
When building with Clang we encounter the following warning:
| drivers/net/wireless/mediatek/mt7601u/eeprom.c:193:5: error: format
| specifies type 'char' but the argument has type 'int' [-Werror,-Wformat]
| chan_bounds[idx].start + chan_bounds[idx].num - 1);
Variadic functions (printf-like) undergo default argument promotion.
Documentation/core-api/printk-formats.rst specifically recommends using
the promoted-to-type's format flag.
Moreover, C11 6.3.1.1 states:
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1548.pdf) `If an int
can represent all values of the original type ..., the value is
converted to an int; otherwise, it is converted to an unsigned int.
These are called the integer promotions.`
With this information in hand, we really should stop using `%hh[dxu]` or
`%h[dxu]` as they usually prompt Clang -Wformat warnings as well as go
against documented standard recommendations.
wifi: p54: Fix an error handling path in p54spi_probe()
If an error occurs after a successful call to p54spi_request_firmware(), it
must be undone by a corresponding release_firmware() as already done in
the error handling path of p54spi_request_firmware() and in the .remove()
function.
Add the missing call in the error handling path and remove it from
p54spi_request_firmware() now that it is the responsibility of the caller
to release the firmware
Johannes Berg [Wed, 1 Jun 2022 19:17:34 +0000 (21:17 +0200)]
wifi: mac80211: support MLO authentication/association with one link
It might seem a bit pointless to do a multi-link operation
connection with just a single link, but this is already a
big change, so for now, limit MLO connections to a single
link.
Extending that to multiple links will require
* work on parsing the multi-link element with STA profile
properly, including element fragmentation;
* checking the per-link status in the multi-link element
* implementing logic to have active/inactive links to let
drivers decide which links should be active;
* implementing multicast RX deduplication;
* and likely more.
For now this is still useful since it lets us do multi-link
connections for the purposes of testing APIs and the higher
layers such as wpa_supplicant.
Johannes Berg [Thu, 30 Jun 2022 11:48:22 +0000 (13:48 +0200)]
wifi: mac80211: add API to parse multi-link element
Add the necessary API to parse the multi-link element in
the future. For now, link only to the element when found
so we can use it in the client-side code later.
Later, we'll need to fill this in to deal with element
fragmentation, parse the STA profile, etc.
Johannes Berg [Thu, 14 Jul 2022 19:58:00 +0000 (21:58 +0200)]
wifi: mac80211_hwsim: fix TX link selection
Now that we have a pointer to the TX STA even when it's
not authenticated/... yet, fix the TX link selection in
hwsim to select only among the valid links for the STA,
requiring a STA pointer here. Also implement a simple
round-robin between links to make life more interesting.
While at it, also consider A3 when translating to link
addresses.
Use the link ID provided in TX frame metadata to select the correct
channel. For now, always select the link with the lowest link ID and
do some address translation.
Johannes Berg [Thu, 14 Jul 2022 21:47:32 +0000 (23:47 +0200)]
wifi: mac80211: do link->MLD address translation on RX
In some cases, e.g. with Qualcomm devices and management
frames, or in hwsim, frames may be reported from the driver
with link addresses, but for decryption and matching needs
we really want to have them with MLD addresses. Support the
translation on RX.
wifi: mac80211: select link when transmitting to non-MLO stations
When an MLO AP is transmitting to a non-MLO station, addr2 should be set
to a link address. This should be done before the frame is encrypted as
otherwise aad verification would fail. In case of software encryption
this can't be left for the device to handle, and should be done by
mac80211 when building the frame hdr.
Johannes Berg [Thu, 14 Jul 2022 21:40:47 +0000 (23:40 +0200)]
wifi: mac80211: fix up link station creation/insertion
When we create a station with a non-default link, then
we should have a link address, and we definitely need
to insert it into the link hash table on insertion.
Split the API into with and without link creation and
if it has a link, insert the link into the link hash
table on sta_info_insert().
Johannes Berg [Thu, 14 Jul 2022 20:23:02 +0000 (22:23 +0200)]
wifi: mac80211_hwsim: implement sta_state for MLO
In MLO, we need to transmit to another MLD and select
the link to it, which requires knowing the station.
But in TX, mac80211 will not give us a station that's
not added to the driver, which in the older add/remove
API is only done later.
Implement the new API in MLO so we know about the STA
at all times and get a pointer during TX as well.
Johannes Berg [Tue, 12 Jul 2022 13:21:23 +0000 (15:21 +0200)]
wifi: mac80211: mlme: refactor assoc link setup
Factor out the code to set up the assoc link into a
new function ieee80211_setup_assoc_link().
While at it, also modify the 'override' handling to
just take into account whether or not the conn_flags
were changed, which is what we need to setup again
the channel later.
Johannes Berg [Tue, 12 Jul 2022 12:48:55 +0000 (14:48 +0200)]
wifi: mac80211: mlme: refactor ieee80211_prep_channel() a bit
Refactor ieee80211_prep_channel() to make the link argument
optional and add a conn_flags pointer argument instead, so
that we can later use this for links that don't exist yet
to build the right information for MLO.
Johannes Berg [Tue, 12 Jul 2022 12:38:02 +0000 (14:38 +0200)]
wifi: mac80211: mlme: refactor assoc req element building
For MLO, we will need to build these elements per link, so
factor out the code that does this, returning the capability,
to simplify building the multi-link element in the future.
Johannes Berg [Tue, 12 Jul 2022 11:46:19 +0000 (13:46 +0200)]
wifi: mac80211: mlme: switch some things back to deflink
With MLO, when we'll disconnect from an AP MLD, we'll just
destroy all the links. Therefore, the only thing we (may)
need to reset is the deflink data, so switch back to that
and adjust the comments accordingly.
Johannes Berg [Tue, 12 Jul 2022 09:33:57 +0000 (11:33 +0200)]
wifi: mac80211: mlme: change flags in ieee80211_determine_chantype()
For MLO we'll need to read flags not directly from the link as
it may not even exist yet if we're just setting up flags for
a secondary link before sending the association request, so
pass the incoming conn_flags separately. Also, while at it,
pass the sdata/link separately as for non-tracking now the
link may be NULL.
Johannes Berg [Tue, 12 Jul 2022 09:13:56 +0000 (11:13 +0200)]
wifi: mac80211: mlme: shift some code around
We'll need ieee80211_prep_channel() in other code for MLO
later, so move the code up - unchanged for now - to avoid
forward declarations in the future.
Johannes Berg [Tue, 12 Jul 2022 15:33:13 +0000 (17:33 +0200)]
wifi: nl80211: fix some attribute policy entries
The new NL80211_CMD_ADD_LINK_STA and NL80211_CMD_MODIFY_LINK_STA
commands have strict policy validation, so fix the policy so it
can be validated correctly.
Johannes Berg [Mon, 11 Jul 2022 12:18:18 +0000 (14:18 +0200)]
wifi: nl80211: reject fragmented and non-inheritance elements
The underlying mac80211 code cannot deal with fragmented
elements for purposes of sorting the elements into the
association frame, so reject those inside the link. We
might want to reject them inside the assoc frame, but
they're used today for FILS, so cannot do that.
The non-inheritance element inside the links similarly
cannot be handled by mac80211, and outside the links it
makes no sense.
Reject both since using them could lead to an incorrect
implementation.
Johannes Berg [Mon, 11 Jul 2022 09:53:20 +0000 (11:53 +0200)]
wifi: nl80211: reject link specific elements on assoc link
When we associate, we'll include all the elements for the
link we're sending the association request on in the frame
and the specific ones for other links in the multi-link
element container. Prohibit adding link-specific elements
for the association link.
Fixes: d648c23024bd ("wifi: nl80211: support MLO in auth/assoc") Signed-off-by: Johannes Berg <[email protected]>
Johannes Berg [Mon, 11 Jul 2022 08:08:11 +0000 (10:08 +0200)]
wifi: nl80211: advertise MLO support
At least while we don't have any more specific interface
combinations support, add a simple flag for MLO support,
we can keep this later based on something other than the
wiphy flag.
wifi: mac80211: Support multi link in ieee80211_recalc_min_chandef()
Recalculate min channel context for the given or all interface
links, depending on the caller. For a station state change, we
need to recalculate all of them since we don't know which link
(or multiple) it might be on.
wifi: mac80211: don't check carrier in chanctx code
We check here that we don't enable TX (netif_carrier_ok())
before we actually start using some channel context, but to
our knowledge this check has never triggered, and with MLO
it's just wrong since links can be added and removed much
more dynamically than before.
Simply remove the checks, there's no really good way to do
anything that would replace them.
wifi: mac80211: Allow EAPOL frames from link addresses
Allow transmitting EAPOL frames not only from the interface
address (which is the MLD address) but also any link addresses,
in order to support non-MLO stations on AP interfaces.
Johannes Berg [Thu, 7 Jul 2022 13:28:14 +0000 (15:28 +0200)]
wifi: mac80211: add a helper to fragment an element
The way this works is that you add all the element data,
keeping a pointer to the length field of the element.
Then call this helper function, which will fragment the
element if there was more than 255 bytes in the element,
memmove()ing the data back if needed.
Johannes Berg [Mon, 4 Jul 2022 13:02:33 +0000 (15:02 +0200)]
wifi: nl80211: add EML/MLD capabilities to per-iftype capabilities
We have the per-interface type capabilities, currently for
extended capabilities, add the EML/MLD capabilities there
to have this advertised by the driver.
Johannes Berg [Wed, 6 Jul 2022 12:03:07 +0000 (14:03 +0200)]
wifi: mac80211: fix link manipulation
When we add non-deflink pointers, we need to remove the
link[0] pointer to deflink in case link[0] is not valid
afterwards. Also, we need to add that back when there
are no more valid links. Reorg the code to fix that.
Johannes Berg [Wed, 6 Jul 2022 07:57:42 +0000 (09:57 +0200)]
wifi: cfg80211: clean up links appropriately
This was missing earlier, we need to remove links when
interfaces are being destroyed, and we also need to
stop (AP) operations when a link is being destroyed.
Address these issues to remove many warnings that will
otherwise appear in mac80211.
Johannes Berg [Fri, 1 Jul 2022 14:14:33 +0000 (16:14 +0200)]
wifi: mac80211: mlme: simplify adding ht/vht/he/eht elements
The functions currently take a link and check data
from it, but this needs to change for MLO. Simplify
the prototypes by passing only the needed arguments.
Remove the regulatory checks, the warnings shouldn't
trigger, and haven't as far as I know.
Johannes Berg [Fri, 1 Jul 2022 13:38:57 +0000 (15:38 +0200)]
wifi: mac80211: refactor adding custom elements
Rework the sorting of custom elements into the association
request by moving the elements before HT/VHT/HE to each
their own function. While at it, fix the placement of the
ones that should be between VHT and HE.
This doesn't fix the placement of elements that should be
between HE and EHT yet, a similar change might be needed
in the future.
Johannes Berg [Thu, 30 Jun 2022 12:20:54 +0000 (14:20 +0200)]
wifi: mac80211: don't set link address for station
We need to handle the link addresses for station differently,
they will be determined by the association code, stored, and
then applied when the links are actually created on success,
cfg80211 will fill in the right addresses per the data we're
sending back to it.
Johannes Berg [Wed, 29 Jun 2022 11:29:05 +0000 (13:29 +0200)]
wifi: mac80211: fix multi-BSSID element parsing
When parsing a frame containing a multi-BSSID element, we
need to know both the transmitted and non-transmitted BSSID
so we can parse it correctly.
Unfortunately, in quite a number of cases, we got this wrong
and were passing the wrong BSSID or useless information:
* the mgmt->bssid from a frame is only the transmitted
BSSID if the frame is a beacon
* passing just one of the parameters as non-NULL isn't
useful and ignored
In those case where we need to parse for a specific BSS we
always have a BSS structure pointer, representing the BSS
we need, whether transmitted or not. Thus, pass that pointer
to the parsing function instead of the two BSSIDs.
Also fix two bugs:
* we need to re-parse all the elements for the other BSS
when iterating the non-transmitted BSSes in scan
* we need to parse for the correct BSS when setting up
the channel data in client code
Fixes: 78ac51f81532 ("mac80211: support multi-bssid") Signed-off-by: Johannes Berg <[email protected]>
projects / linux.git / log