John Baldwin [Tue, 14 Dec 2021 19:01:05 +0000 (11:01 -0800)]
ktls: Support for TLS 1.3 receive offload.
Note that support for TLS 1.3 receive offload in OpenSSL is still an
open pull request in active development. However, potential changes
to that pull request should not affect the kernel interface.
John Baldwin [Thu, 9 Dec 2021 21:17:54 +0000 (13:17 -0800)]
TLS: Use <machine/tls.h> for libc and rtld.
- Include <machine/tls.h> in MD rtld_machdep.h headers.
- Remove local definitions of TLS_* constants from rtld_machdep.h
headers and libc using the values from <machine/tls.h> instead.
- Use _tcb_set() instead of inlined versions in MD
allocate_initial_tls() routines in rtld. The one exception is amd64
whose _tcb_set() invokes the amd64_set_fsbase ifunc. rtld cannot
use ifuncs, so amd64 inlines the logic to optionally write to fsbase
directly.
- Use _tcb_set() instead of _set_tp() in libc.
- Use '&_tcb_get()->tcb_dtv' instead of _get_tp() in both rtld and libc.
This permits removing _get_tp.c from rtld.
- Use TLS_TCB_SIZE and TLS_TCB_ALIGN with allocate_tls() in MD
allocate_initial_tls() routines in rtld.
John Baldwin [Thu, 9 Dec 2021 21:17:41 +0000 (13:17 -0800)]
libthr: Use <machine/tls.h> for most MD TLS details.
Note that on amd64 this effectively removes the unused tcb_spare field
from the end of struct tcb since the definition of struct tcb in
<x86/tls.h> does not include that field.
Reviewed by: kib, jrtc27
Sponsored by: The University of Cambridge, Google Inc.
Differential Revision: https://reviews.freebsd.org/D33352
John Baldwin [Thu, 9 Dec 2021 21:17:13 +0000 (13:17 -0800)]
Add <machine/tls.h> header to hold MD constants and helpers for TLS.
The header exports the following:
- Definition of struct tcb.
- Helpers to get/set the tcb for the current thread.
- TLS_TCB_SIZE (size of TCB)
- TLS_TCB_ALIGN (alignment of TCB)
- TLS_VARIANT_I or TLS_VARIANT_II
- TLS_DTV_OFFSET (bias of pointers in dtv[])
- TLS_TP_OFFSET (bias of "thread pointer" relative to TCB)
Note that TLS_TP_OFFSET does not account for if the unbiased thread
pointer points to the start of the TCB (arm and x86) or the end of the
TCB (MIPS, PowerPC, and RISC-V).
Note also that for amd64, the struct tcb does not include the unused
tcb_spare field included in the current structure in libthr. libthr
does not use this field, and the existing calls in libc and rtld that
allocate a TCB for amd64 assume it is the size of 3 Elf_Addr's (and
thus do not allocate room for tcb_spare).
A <sys/_tls_variant_i.h> header is used by architectures using
Variant I TLS which uses a common struct tcb.
Reviewed by: kib (older version of x86/tls.h), jrtc27
Sponsored by: The University of Cambridge, Google Inc.
Differential Revision: https://reviews.freebsd.org/D33351
For stable/13 only, sys/arm/include/tls.h includes support for
ARM_TP_ADDRESS which is not present in main.
John Baldwin [Thu, 9 Dec 2021 21:16:19 +0000 (13:16 -0800)]
mips: Add TLS_DTV_OFFSET to the result of tls_get_addr_common.
Previously TLS_DTV_OFFSET was added to the offset passed to
tls_get_addr_common; however, this approach matches powerpc and RISC-V
and better matches the intention.
Reviewed by: kib, jrtc27
Sponsored by: The University of Cambridge, Google Inc.
Differential Revision: https://reviews.freebsd.org/D33347
John Baldwin [Thu, 9 Dec 2021 21:16:00 +0000 (13:16 -0800)]
mips: Rename TLS_DTP_OFFSET to TLS_DTV_OFFSET.
This is the more standard name for the bias of dtv pointers used on
other platforms. This also fixes a few other places that were using
the wrong bias previously on MIPS such as dlpi_tls_data in struct
dl_phdr_info and the recently added __libc_tls_get_addr().
Reviewed by: kib, jrtc27
Sponsored by: The University of Cambridge, Google Inc.
Differential Revision: https://reviews.freebsd.org/D33346
John Baldwin [Thu, 9 Dec 2021 21:15:38 +0000 (13:15 -0800)]
libthr: Remove the DTV_OFFSET macro.
This macro is confusing as it is not related to the similarly named
TLS_DTV_OFFSET. Instead, replace its one use with the desired
expression which is the same on all platforms.
Reviewed by: kib, emaste, jrtc27
Sponsored by: The University of Cambridge, Google Inc.
Differential Revision: https://reviews.freebsd.org/D33345
John Baldwin [Thu, 9 Dec 2021 19:52:42 +0000 (11:52 -0800)]
GMAC: Reset initial hash value and counter in AES_GMAC_Reinit().
Previously, these values were only cleared in AES_GMAC_Init(), so a
second set of operations could reuse the final hash as the initial
hash. Currently this bug does not trigger in cryptosoft as existing
GMAC and GCM operations always use an on-stack auth context
initialized from a template context.
Reviewed by: markj
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D33315
John Baldwin [Thu, 9 Dec 2021 19:52:41 +0000 (11:52 -0800)]
crypto: Don't assert for empty output buffers.
It is always valid for crp_payload_output_start to be 0. However, if
an output buffer is empty (e.g. a decryption request with a tag but an
empty payload), the existing assertion failed since 0 is not less than
0.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D33193
John Baldwin [Fri, 30 Jul 2021 00:09:23 +0000 (17:09 -0700)]
geom_vfs: Pre-allocate event for g_vfs_destroy.
When an active g_vfs is orphaned due to an underlying disk going away
the destroy is deferred until the filesystem is unmounted in
g_vfs_done(). However, g_vfs_done() is invoked from a non-sleepable
context and cannot use M_WAITOK to allocate the event. Instead,
allocate the event in g_vfs_orphan() and save it in the softc to be
retrieved by the last call to g_vfs_done().
Alan Somers [Sat, 2 Apr 2022 19:31:24 +0000 (13:31 -0600)]
fusefs: fix two bugs regarding VOP_RECLAIM of the root inode
* We never send FUSE_LOOKUP for the root inode, since its inode number
is hard-coded to 1. Therefore, we should not send FUSE_FORGET for it,
lest the server see its lookup count fall below 0.
* During VOP_RECLAIM, if we are reclaiming the root inode, we must clear
the file system's vroot pointer. Otherwise it will be left pointing
at a reclaimed vnode, which will cause future VOP_LOOKUP operations to
fail. Previously we only cleared that pointer during VFS_UMOUNT. I
don't know of any real-world way to trigger this bug.
Support for this directive has been removed in config(8) on main,
which leaves us unable to build LINT with newer config(8). It's
believed that mcount-based profiling didn't really work on modern
systems anyways, so the value of testing this is low.
We avoid providing limited backwards compatibility here to continue and
warn folks that may somehow be deploying real-world configs with `profile`
specified.
This is a direct commit to stable/13, but a partial MFC of aa3ea612be36.
Mark Johnston [Thu, 21 Apr 2022 14:49:22 +0000 (10:49 -0400)]
ctfdump: Remove definitions of warn() and vwarn()
The presence of the latter causes a link error when building a
statically linked ctfdump(1) because libc defines the same symbol.
libc's warn() is defined as a weak symbol and so does not cause the same
problem, but let's just use libc's version.
Reported by: stephane rochoy <stephane.rochoy@stormshield.eu>
Sponsored by: The FreeBSD Foundation
Mark Johnston [Thu, 21 Apr 2022 17:22:09 +0000 (13:22 -0400)]
mld6: Ensure that mld_domifattach() always succeeds
mld_domifattach() does a memory allocation under the global MLD mutex
and so can fail, but no error handling prevents a null pointer
dereference in this case. The mutex is only needed when updating the
global softc list; the allocation and static initialization of the softc
does not require this mutex. So, reduce the scope of the mutex and use
M_WAITOK for the allocation.
xhci(4): Ensure the so-called data toggle gets properly reset.
Use the drop and enable endpoint context commands to force a reset of
the data toggle for USB 2.0 and USB 3.0 after:
- clear endpoint halt command (when the driver wishes).
- set config command (when the kernel or user-space wants).
- set alternate setting command (only affected endpoints).
Some XHCI HW implementations may not allow the endpoint reset command when
the endpoint context is not in the halted state.
Reported by: Juniper and Gary Jennejohn
Sponsored by: NVIDIA Networking
e1000: Try auto-negotiation for fixed 100 or 10 configuration
Currently if an e1000 interface is set to a fixed media configuration,
for gigabit, it will participate in auto-negotiation as required by
IEEE 802.3-2018 Clause 37. However, if set to fixed media configuration
for 100 or 10, it does NOT participate in auto-negotiation.
By my reading of Clauses 28 and 37, while auto-negotiation is optional
for 100 and 10, it is not prohibited and is, in fact, "highly
recommended".
This patch enables auto-negotiation for fixed 100 and 10 media
configuration, in a similar manner to that already performed for 1000.
I.e., the patch enables advertising of just the manually configured
settings with the goal of allowing the remote end to match the manually
configured settings if it has them available.
To be clear, this patch does NOT allow an em(4) interface that has been
manually configured with specific media settings to respond to
auto-negotiation by then configuring different parameters to those that
were manually configured. The intent of this patch is to fully comply
with the requirements of Clause 37, but for 100 and 10.
The need for this has arisen on an em(4) link where the other end is
under a different administrative control and is set to full
auto-negotiation. Due to the cable length GigE is not working well. It
is desired to set the em(4) end to "media 100baseTX mediatype
full-duplex" which does work when both ends are configured that way.
Currently, because em(4) does not participate in autoneg for this
setting, the remote defaults to half-duplex - i.e., there's a duplex
mismatch and things don't work. With this patch, em(4) would inform the
remote that it has only 100baseTX full, the remote would match that and
it will work.
Ed Maste [Wed, 27 Apr 2022 13:15:09 +0000 (09:15 -0400)]
libcxxrt: Insert padding in __cxa_dependent_exception
Padding was added to __cxa_exception in 45ca8b19 and
__cxa_dependent_exception needs the same layout.
Add some static_asserts to detect this in the future.
John F. Carr [Sat, 19 Mar 2022 22:51:43 +0000 (18:51 -0400)]
hpet: Allow a MMIO window smaller than 1K
Some new AMD systems provide a HPET MMIO region smaller than the 1KB
specified, and a correspondingly small number of timers. Handle this in
the HPET driver rather than requiring a 1KB window. This allows the
HPET driver to attach on such systems.
pf: counter argument to pfr_pool_get() may never be NULL
Coverity points out that if counter was NULL when passed to
pfr_pool_get() we could potentially end up dereferencing it.
Happily all users of the function pass a non-NULL pointer. Enforce this
by assertion and remove the pointless NULL check.
Move the use of 'sc' to after the NULL check.
It's very unlikely that we'd actually hit this, but Coverity is correct
that it's not a good idea to dereference the pointer and only then NULL
check it.
15b1eb142c changed the callout code to store the CALLOUT_SHAREDLOCK flag
in c_iflags (where it used to be c_flags), but failed to update the
check in softclock_call_cc(). This resulted in the callout code always
taking the write lock, even if a read lock had been requested (with
the CALLOUT_SHAREDLOCK flag in callout_init_rm()).
Ed Maste [Tue, 19 Apr 2022 19:44:46 +0000 (15:44 -0400)]
capsicum: briefly describe capabilities in man page
Provide a very brief introduction to capabilities, using a couple of
sentences from David Chisnall's mailing list response[1] to a question
about Linux capabilities and Capsicum.
Mailing list subject (in case the archive URL changes) was
Re: Linux capabilities to Capsicum
Kyle Evans [Sat, 19 Mar 2022 03:03:44 +0000 (22:03 -0500)]
psci: finish psci_present implementation
This was already declared in psci.h, but it was never defined/set. Do
this now, so we can use it to decide if enable-method in /cpus FDT nodes
should be inspected later on. While we're here, convert it to a
boolean.
stand: zfs: handle holes at the tail end correctly
This mirrors dmu_read_impl(), zeroing out the tail end of the buffer and
clipping the read to what's contained by the block that exists.
This fixes an issue that arose during the 13.1 release process; in
13.1-RC1 and later, setting up GELI+ZFS will result in a failure to
boot. The culprit is this, which causes us to fail to load geom_eli.ko
as there's a residual portion after the single datablk that should be
zeroed out.
The correct logic is a lot simpler than the previous iteration. We
record the base fts_name to avoid having to worry about whether we
needed the root symlink name or not (as applicable), then we can simply
shift all of that logic to after path translation to make it less
fragile.
If we're copying to DNE, then we'll have swapped out the NULL root_stat
pointer and then attempted to recurse on it. The previously nonexistent
directory shouldn't exist at all in the new structure, so just back out
from that tree entirely and move on.
The tests have been amended to indicate our expectations better with
subdirectory recursion. If we copy A to A/B, then we expect to copy
everything from A/B/* into A/B/A/B, with exception to the A that we
create in A/B.
Kyle Evans [Thu, 27 Jan 2022 18:02:17 +0000 (12:02 -0600)]
cp: fix some cases with infinite recursion
As noted in the PR, cp -R has some surprising behavior. Typically, when
you `cp -R foo bar` where both foo and bar exist, foo is cleanly copied
to foo/bar. When you `cp -R foo foo` (where foo clearly exists), cp(1)
goes a little off the rails as it creates foo/foo, then discovers that
and creates foo/foo/foo, so on and so forth, until it eventually fails.
POSIX doesn't seem to disallow this behavior, but it isn't very useful.
GNU cp(1) will detect the recursion and squash it, but emit a message in
the process that it has done so.
This change seemingly follows the GNU behavior, but it currently doesn't
warn about the situation -- the author feels that the final product is
about what one might expect from doing this and thus, doesn't need a
warning. The author doesn't feel strongly about this.
PR: 235438
Reviewed by: bapt
Sponsored by: Klara, Inc.
Wojciech Macek [Mon, 20 Dec 2021 05:32:51 +0000 (06:32 +0100)]
cam: don't send scsi commands on shutdown when reboot method RB_NOSYNC
Don't send the SCSI comand SYNCHRONIZE CACHE on devices that are still
open when RB_NOSYNC is the reboot method. This may avoid recursive panics
when doadump is called due to a SCSI/CAM/USB error/bug.
Rick Macklem [Sun, 23 Jan 2022 22:17:40 +0000 (14:17 -0800)]
mountd: Delay starting mountd until after mountlate
PR#254282 reports a problem where nullfs mounts cannot be
exported via mountd for FreeBSD 13.0.
The problem seems to be that, to do the nullfs mounts in
/etc/fstab, they require the "late" mount option, so that the
underlying filesystem is mounted (ZFS for the PR).
Adding "mountlate" to the REQUIRE list in /etc/rc.d/mountd
fixes the problem, but that results in a dependency cycle
because /etc/rc.d/lockd specifies:
REQUIRE: nfsd
BEFORE: DAEMON
--> which forces mountd to preceed DAEMON.
This patch removes "nfsd" from REQUIRE for lockd and statd,
then adds mountlate to REQUIRE for mountd, to fix this
problem. Having lockd REQUIRE nfsd was done in the NetBSD
code when it was pulled into FreeBSD and there does not
seem to be a need for this.
In case this causes problems, a long MFC has been specified.
Mark Johnston [Wed, 13 Apr 2022 14:47:08 +0000 (10:47 -0400)]
libsysdecode: Fix decoding of Capsicum rights
Capsicum rights are a bit tricky since some of them are subsets of
others, and one can have rights R1 and R2 such that R1 is a subset of
R2, but there is no collection of named rights whose union is R2. So,
they don't behave like most other flag sets. sysdecode_cap_rights(3)
does not handle this properly and so can emit misleading decodings.
Try to fix all of these problems:
- Include composite rights in the caprights table.
- Use a constructor to sort the caprights table such that "larger"
rights appear first and thus are matched first.
- Don't print rights that are a subset of rights already printed, so as
to minimize the length of the output.
- Print a trailing message if some of the specific rights are not
matched by the table.
PR: 263165
Reviewed by: pauamma_gundo.com (doc), jhb, emaste
Sponsored by: The FreeBSD Foundation
Mark Johnston [Wed, 23 Mar 2022 16:36:12 +0000 (12:36 -0400)]
setitimer: Fix exit race
We use the p_itcallout callout, interlocked by the proc lock, to
schedule timeouts for the setitimer(2) system call. When a process
exits, the callout must be stopped before the process struct is
recycled.
Currently we attempt to stop the callout in exit1() with the call
_callout_stop_safe(&p->p_itcallout, CS_EXECUTING). If this call returns
0, then we sleep in order to drain the callout. However, this happens
only if the callout is not scheduled at all. If the callout thread is
blocked on the proc lock, then exit1() will not block and the callout
may execute after the process has fully exited, typically resulting in a
panic.
I cannot see a reason to use the CS_EXECUTING flag here. Instead, use
the regular callout_stop()/callout_drain() dance to halt the callout.
Reported by: ler
Tested by: ler, pho
Sponsored by: The FreeBSD Foundation
diff: tests: loosen up requirements for report_identical
This test cannot run without an unprivileged_user being specified
anyways, so just run as the unprivileged user. Revoking read permisions
works just as well if you're guaranteed non-root.
loader: userboot: provide a getsecs() implementation
We don't need it for userboot, but it avoids issues with BIND_NOW, so
just provide it. time(3) isn't defined but ends up being provided by
libc linked into the host process, which is generally fine.
Printing device followed by interface matches, e.g., edk2. Note that
this is only a fallback, many firmware implementations will provide the
protocol that we'll use to format device paths.
Mike Karels [Mon, 11 Apr 2022 19:44:49 +0000 (14:44 -0500)]
genet: fix problems with interface down/up
The genet interface did not resume operation correctly after doing
ifconfig down then up. The down/reset procedure did not clear the
RUNNING flag, and did not reset enough of the hardware state. This
patch is modeled on OpenBSD code, with a call to gen_reset added
to reset the controller completely. Regularize the parameter to
gen_dma_disable() while here.
Mark Johnston [Thu, 14 Apr 2022 19:45:54 +0000 (15:45 -0400)]
vm: Move the "vm_wait in early boot" assertion to the proper place
The assertion was added in commit 1771e987ca6a. After that, vm_wait()
and friends were refactored such that the actual sleep happens
elsewhere. Now the assertion condition is not checked when
vm_wait_doms() is called directly, and it is checked even if we are not
going to sleep (because vm_page_count_min_set(wdoms) is false).
Reviewed by: alc, kib
Sponsored by: The FreeBSD Foundation
Gordon Bergling [Thu, 14 Apr 2022 08:04:14 +0000 (10:04 +0200)]
time(3): Refine history in the manual page
The time() system call first appeared in Version 1 AT&T UNIX. Through
the Version 3 AT&T UNIX, it returned 60 Hz ticks since an epoch that
changed occasionally, because it was a 32-bit value that overflowed in a
little over 2 years.
In Version 4 AT&T UNIX the granularity of the return value was reduced to
whole seconds, delaying the aforementioned overflow until 2038.
Version 7 AT&T UNIX introduced the ftime() system call, which returned
time at a millisecond level, though retained the gtime() system call
(exposed as time() in userland). time() could have been implemented as a
wrapper around ftime(), but that wasn't done.
4.1cBSD implemented a higher-precision time function gettimeofday() to
replace ftime() and reimplemented time() in terms of that.
Since FreeBSD 9 the implementation of time() uses
clock_gettime(CLOCK_SECOND) instead of gettimeofday() for performance
reasons.
Mark Johnston [Mon, 18 Apr 2022 15:45:45 +0000 (11:45 -0400)]
path_test: Correct the kevent test
Perhaps surprisingly, and contrary to the expectations of
path_test:path_event, NOTE_LINK events are not raised when a file is
unlinked. Prior to commit bf13db086b84, the test happened to work
because unlinking the file would cause the vnode to be recycled, and
EVFILT_VNODE knotes deliver an event with EV_EOF set when the vnode is
doomed. Since the test did not verify the note type, the test
succeeded. After commit bf13db086b84, the vnode is not recycled after
being unlinked and so the test hangs.
Fix the test by waiting for NOTE_DELETE instead, and check that we got
the note that we expected.
Reported by: Jenkins
Sponsored by: The FreeBSD Foundation
Insert padding in __cxa_exception struct for compatibility
Similar to https://github.com/llvm/llvm-project/commit/f2a436058fcb, the
addition of __attribute__((__aligned__)) to _Unwind_Exception (in commit b9616964) causes implicit padding to be inserted before the unwindHeader
field in __cxa_exception.
Applications attempt to get at the earlier fields in __cxa_exception, so
preserve the same negative offsets in __cxa_exception, by moving the
padding to the beginning of the struct.
The assumption here is that if the ABI is not aware of the padding
before unwindHeader and put the referenceCount/primaryException in
there, no padding should exist before unwindHeader.
This should make libreoffice's custom exception handling mechanisms work
correctly, even if it was built against an older cxxabi.h/unwind.h pair.
PR: 263370
Approved by: re (gjb, early MFC)
MFC after: immediately
Andrew Turner [Tue, 22 Mar 2022 15:46:15 +0000 (15:46 +0000)]
Add an implementation of .mcount on arm64
To support cc -pg on arm64 we need to implement .mcount. As clang and
gcc think it is function like it just needs to load the arguments
to _mcount and call it.
On gcc the first argument is passed in x0, however this is missing on
clang so we need to load it from the stack. As it's the caller return
address this will be at a known location.
PR: 262709
Reviewed by: emaste (earlier version)
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D34634
Andrew Turner [Fri, 5 Feb 2021 10:50:29 +0000 (10:50 +0000)]
Add support for arm64 nGnRE device memory
On arm64 we can select how strongly we order device memory. Currently
we use the strongest type of non-Gathering, non-Reordering, no Early
write acknowledgement. This is equivalent to VM_MEMATTR_SO in the 32-bit
arm code.
Create a new memory type to remove the no Early write acknowledgement
option to create a memory attribute that is equivalent to the arm
VM_MEMATTR_DEVICE.
Keep the the old nGnRnE memory as what we provide for VM_MEMATTR_DEVICE
until we can test nGnRE on more hardware. A method for dynamically
switching back may be needed as at least one vendor is known to have
broken nGnRE memory.
This code was marked gone_in(13), so its time has passed.
The only consumer of this interface is dumpon(8). We do not maintain
strict backwards compatibility for this utility because a) it
can't/shouldn't be used from a jail or chroot and b) it is highly
specific interface unique to FreeBSD. The host's (presumably more
up-to-date) copy of dumpon(8) should be used to configure kernel dump
devices.
LinuxKPI: 802.11: start adding rate control to ieee80211_tx_status()
Start adding rate control feedback in ieee80211_tx_status() in order
for net80211 to be able to report something back (which may not
yet be the view of the firmware). iwlwifi is reporting back an MSC 0
even with HT disabled (to be investigated) so we cannot (yet) use
the firmware/driver rate feedback directly.
Implement skb_copy() with omissions of fragments and possibly other fields
for now. Should we hit frags at any point a log message will let us know.
For the few cases we need this currently this is enough.
Initially we were using the IEs from ieee80211_probereq_ie() of net80211
and put them into the common_ies field. Start by manually building the
per-band and common IE parts as drivers put them back together.
This also involves allocating the req.ie as one buffer for all IEs over
all bands and setting req.ie_len correctly based on how many bytes we
put in.
Manually building per-band scan IEs we still use the net80211 routines
to add IEs to the buffer (mostly).
This is needed by Realtek drivers but will equally used by others.
Realtek would simply panic due to skbs being allocated with the wrong
length.
Longer-term this will help us, e.g., when not supporting VHT on 2Ghz
and we would have to do this anyway.
LinuxKPI: skbuff: dev_kfree_skb_irq() and improvements
While it is currently unclear if we will have to defer work in
dev_kfree_skb_irq() to call dev_kfree_skb().
We only have one caller which seems to be fine on FreeBSD by calling
it directly for now.
While here shortcut skb_put()/skb_put_data() saving us work if there
are no adjustments to do.
Also adjust the logging in skb_is_gso() to avoid getting spammed by it.
LinuxKPI: 802.11: use an sx lock to protect the list of vifs
Use an sx lock to protect the list of vifs. We could use the
linux mutex compat for this but our current implementation may
re-acquire the lock recursively so allow this. The change is
mainly motivated by the fact that some callers may sleep in the
interator function called. Recursiveness is needed because we
see find_sta_by_ifaddr() being called from an iterator function
from iterate_interfaces().