Currently softnet_lock is taken unnecessarily in some cases, e.g.,
icmp_input and encap4_input from ip_input, or not taken even if needed,
e.g., udp_input and tcp_input from ipsec4_common_input_cb. Fix them.
NFC if NET_MPSAFE is disabled (default).
A typical ussage of localcount_drain looks like this:
mutex_enter(&mtx);
item = remove_from_list();
pserialize_perform(psz);
localcount_drain(&item->localcount, &cv, &mtx);
mutex_exit(&mtx);
This sequence can cause a deadlock which happens for example on the following
situation:
- Thread A calls localcount_drain which calls xc_broadcast after releasing
a specified mutex
- Thread B enters the sequence and calls pserialize_perform with holding
the mutex while pserialize_perform also calls xc_broadcast
- Thread C (xc_thread) that calls an xcall callback of localcount_drain tries
to hold the mutex
xc_broadcast of thread B doesn't start until xc_broadcast of thread A
finishes, which is a feature of xcall(9). This means that pserialize_perform
never complete until xc_broadcast of thread A finishes. On the other hand,
thread C that is a callee of xc_broadcast of thread A sticks on the mutex.
Finally the threads block each other (A blocks B, B blocks C and C blocks A).
A possible fix is to serialize executions of the above sequence by another
mutex, but adding another mutex makes the code complex, so fix the deadlock
by another way; the fix is to release the mutex before pserialize_perform
and instead use a condvar to prevent pserialize_perform from being called
simultaneously.
Note that the deadlock has happened only if NET_MPSAFE is enabled.
https://www.intel.com/content/www/us/en/standards/
high-definition-audio-specification.html
page 186 shows the layout of the baseline block of the ELD (EDID Like Data)
struct - and allows a reserved (effectively padding) area at the end of the
struct. This is required to keep the struct an even number of words long
(size measured in units of 32 bits) while allowing for a variable length
monitor name, followed by a variable number of 3 byte structs - the
combination of which is not likely to be a multiple of 4.
Code here assumed that there was no padding, and objected to the ELD
format if any padding bytes existed (hdafg_dd_parse_info() would return
EINVAL) causing a "failed to parse ELD data" message (if HDAFG_HDMI_DEBUG
is defined) from hdafg_assoc_dump_dd() making it difficult (or at least
confusing) to debug HDMI related audio issues (hdafg_assoc_dump_dd would
not print most of the data it is expected to print) although this would
most likely have no effect on actual operations.
Change a test from a != to < (there must be enough data, not exactly the
amount needed) for the EINVAL. As a consequence, the length after the
SAD data is parsed (the 3 byte structs) is no longer required to be 0,
so remove the KASSERT() (previously it was just useless, the code guaranteed
a 0 value, now it is incorrect.) While here also change a related
diagnostic message to be slightly more informative as to what is being shown.
OK jmcneill@
"EERD polling failed" error on attach. if_wm.c rev. 1.534 was not perfect.
Reported by ozaki-r@.
- Print NVM offset and word count when EERD polling failed.
XXX pullup-8.
When we create a pipe to read data, restore the sigchld signal handler since
pclose expects to wait for the child (to avoid spurious error reporting later).
keysock now protects itself by its own mutex, which means that
the rawcb list is protected by two different mutexes (keysock's one
and softnet_lock for rtsock), of course it's useless.
Fix the situation by having a discrete rawcb list for each.
between when xhci interrupts are enabled and the usb bus driver is
attached. If an irq occurs in this window, xhci will attempt to schedule
a softint with an invalid softint handle.
Add a quirk flag, XHCI_DEFERRED_START, that when set skips starting the
controller at the end of xhci_init. Bus glue that sets this is responsible
to call xhci_start after attaching the child usb devices.
