vr(4): set or clear VR_RXCFG_RX_PROMISC bit in vr_setmulti().
Previously the promiscuous bit was only handled in vr_init().
When an interface is added to or removed from bridge(4), its
IFF_PROMISC flag is changed through the interface flags (SIOCSIFFLAGS).
On a running interface it does not reinitialize device, but calls
vr_setmulti() directly, which did not update the promiscuous bit in RX filter.
PR port-i386/60884
printf(3): use emphasis when a term is used for the first time
Use it for "conversion specification", "flag", "field width,
"precision", "length modifier", "conversion specifier".
printf(3): tweak formatting of n$
XXX: The way the description of positional arguments is worded is
pretty awkward. It [ab]uses "field" to refer to the conversion
specification itself, when the standard already uses it to refer to
the output field.
Pull up following revision(s) (requested by kre in ticket #1364):
common/lib/libc/string/memset2.c: revision 1.12
PR port-mips/60780 UBSan: long int << fix (src/common/lib/libc/string/memset2.c)
The comments say that memword_t is intended to be unsigned, but
presume that __register_t will provide that attribute. It usually
doesn't, so we need an unsigned type for building the fill value.
Thanks to Alexander Schreiber for running the UBSan built system
and reporting this (and others), and hopefully, in advance for
testing that this one is now fixed.
Pull up following revision(s) (requested by kre in ticket #1363):
lib/libc/stdio/vfwprintf.c: revision 1.42,1.43
PR lib/60708 - include 'F' format when fetching %n$ args.
PR bin/60496 - fix printf(3) with non-single-byte decimal point & thousands sep
These fixes are relatively simple, and should be safe -- I have had them
in my source tree for ages now.
The fix for the decimal point case is taken more or less directly from
an old FreeBSD fix for the same issue.
The thousands separator case is different - our code and FreeBSD's have
diverged so much in this area that simply using their method would have
been more trouble than it was worth, so I reimplemented it differently
to better fit our code model.
[3 lines not shown]
Pull up following revision(s) (requested by kre in ticket #531):
lib/libc/stdio/vfwprintf.c: revision 1.42,1.43
PR lib/60708 - include 'F' format when fetching %n$ args.
PR bin/60496 - fix printf(3) with non-single-byte decimal point & thousands sep
These fixes are relatively simple, and should be safe -- I have had them
in my source tree for ages now.
The fix for the decimal point case is taken more or less directly from
an old FreeBSD fix for the same issue.
The thousands separator case is different - our code and FreeBSD's have
diverged so much in this area that simply using their method would have
been more trouble than it was worth, so I reimplemented it differently
to better fit our code model.
[3 lines not shown]
Pull up following revision(s) (requested by nat in ticket #530):
sys/arch/mac68k/obio/ascaudio.c: revision 1.19
Return early if aux.interrupts.
Patch and testing by Ray Tran.
Pull up following revision(s) (requested by nat in ticket #529):
sys/arch/mac68k/obio/ascaudio.c: revision 1.20
Honor playback block size within resonable limits.
Patch and testing by Ray Tran.
Pull up following revision(s) (requested by nat in ticket #528):
sys/arch/mac68k/obio/ascaudio.c: revision 1.21
Fix typo for building ascaudio(4) as a module.
Patch and testing by Ray Tran.
Pull up following revision(s) (requested by nat in ticket #527):
sys/arch/mac68k/obio/ascaudio.c: revision 1.22
Respect maximum and minimum values when recording.
Patch and testing by Ray Tran.
Pull up following revision(s) (requested by nat in ticket #526):
sys/arch/mac68k/obio/ascaudio.c: revision 1.23
Avoid use of lock before inited.
Patch and testing by Ray Tran.
Pull up following revision(s) (requested by nat in ticket #525):
sys/arch/mac68k/dev/akbd.c: revision 1.31
Honor keyboard bell volume.
Addresses PR port-mac68k/60691.
Patch and testing by Ray Tran.
Pull up following revision(s) (requested by mrg in ticket #1362):
sys/net/bsd-comp.c: revision 1.24
Properly bounds check the max_ent check in bsd_decompress().
A specially crafted packet from PPP can trigger a backwards write of
an mbuf of a character, and can cause hangs and crashes as it corrupts
memory in the kernel.
The check against max_ent has an off-by-one error, fixed in this change
and fixes the problem (tested on arm64.)
From Acts1631 via OpenBSD.
Pull up following revision(s) (requested by mrg in ticket #524):
sys/net/bsd-comp.c: revision 1.24
Properly bounds check the max_ent check in bsd_decompress().
A specially crafted packet from PPP can trigger a backwards write of
an mbuf of a character, and can cause hangs and crashes as it corrupts
memory in the kernel.
The check against max_ent has an off-by-one error, fixed in this change
and fixes the problem (tested on arm64.)
From Acts1631 via OpenBSD.
Pull up following revision(s) (requested by kre in ticket #523):
common/lib/libc/string/memset2.c: revision 1.12
PR port-mips/60780 UBSan: long int << fix (src/common/lib/libc/string/memset2.c)
The comments say that memword_t is intended to be unsigned, but
presume that __register_t will provide that attribute. It usually
doesn't, so we need an unsigned type for building the fill value.
Thanks to Alexander Schreiber for running the UBSan built system
and reporting this (and others), and hopefully, in advance for
testing that this one is now fixed.
Pull up following revision(s) (requested by tls in ticket #521):
sys/arch/hppa/spmath/Makefile: revision 1.7
sys/arch/hppa/spmath/Makefile.inc: revision 1.13
Use tool tsort, not host tsort, for hppa spmath build
Pull up following revision(s) (requested by andvar in ticket #1361):
sys/dev/pci/if_vr.c: revision 1.139
vr(4): reinitialize device when rx engine fails to go to the idle state.
When detecting root via NFS, DHCP timed out on some boots.
Investigation showed that in certain cases RX engine was getting stuck,
likely triggered by an ARP frame arriving before driver handled a link up
state.
NFS boot's RARP fallback was calling media up command, triggering
reinitialization and forced reset, restoring the device state.
Since fallback doesn't set the IP address, NFS boot failed,
requesting to prompt root device again.
Subsequent attempt succeeded since device was reinitialized
[12 lines not shown]
