On 2019-09-01, Christian Weisgerber wrote:
> This removes the support for building kernels and boot loaders with
> gcc on amd64 and i386, simplifying the corresponding Makefiles.
daniel@ has objected directly to me that he'd like to continue to
be able to build the tree with po
This removes the support for building kernels and boot loaders with
gcc on amd64 and i386, simplifying the corresponding Makefiles.
ok?
Index: arch/amd64/conf/Makefile.amd64
===
RCS file: /cvs/src/sys/arch/amd64/conf/Makefile.amd64,v
On 2019-08-14, Christian Weisgerber wrote:
> clean_old() {
> local _cver
>
> rm -rf /mnt/usr/lib/gcc-lib/!(*$VNAME)
> _cver=$(cd /mnt/usr/lib/clang 2>/dev/null && ls -r | sed -e 1q) &&
> rm -rf /mnt/usr/lib/clan
On 2019-08-13, "Theo de Raadt" wrote:
>> Could you add '&& -d /mnt/usr/lib/clang' to check whether the directory
>> exists before trying to cd into it (similarly for /mnt/usr/bin/gcc and
>> gcc-libs)?
>
> Then might as well just -d only, and skip the -f
Seriously, why bother with a check in the
* Remove syspatch files from the installed system and not the ramdisk.
* Use extended globs and generally adopt to the style of this script.
ok?
I'm not very happy with the way the clang version is determined.
If we ever were to move to 10.0.0, this would remove the wrong
directory. I considered
On 2019-08-08, Stuart Henderson wrote:
> Works as expected for IPv4. For IPv6 it only works if you use a hostname that
> only has a v6 address and not also a v4 address.
Not sure what scenario you mean exactly. This works fine for me:
openrsync --address=2003:e5:cf15:c101:76dd:6f32:911c:ce5c -
Seen on misc@:
> $ ssh-keygen -t ed25519 -b 2
> key bits exceeds maximum 16384
The check for > OPENSSL_RSA_MAX_MODULUS_BITS should only be applied
to RSA keys. No point in checking for > OPENSSL_DSA_MAX_MODULUS_BITS,
since we only permit 1024-bit DSA keys anyway.
While there, only set DEFAU
Chris Ross:
> Is there a hard reason SR_BOOT_LOADER_SIZE cannot be raised to sometime
> larger, perhaps 1024 (instead of the current 320) ?
That size is part of the disk layout.
Looking at , the layout of a softraid partition
is this:
16 sectors, unused (potential disklabel, boot block)
w...@disroot.org:
> after locate(1)'ing and adding those files i could happily run cvsweb on
> openbsd 6.5-release
>
> i'm sending patch for README (2 lines) as suggested by \renaud on #openbsd
> (freenode)
Thanks, but this was already fixed in -current seven weeks ago.
--
Christian "naddy"
Björn Ketelaars:
> Diff below is based on the latest diff from naddy@. Changes:
> - reallocarray likes type_t, as such changes type of nxdev and i;
> - use reallocarray instead of malloc as xdev is initialised as NULL.
ok naddy@
--
Christian "naddy" Weisgerber na...@mip
Christian Weisgerber:
> > To reproduce:
> >
> > mkdir -p /tmp/test /tmp/plop
> > openrsync -rx /tmp/test/ /tmp/plop/
> >
> > Result:
> >
> > openrsync(3470) in free(): bogus pointer (double free?) 0x7f7dcdc8
> > Abort tr
Hiltjo Posthuma:
> I noticed a free() issue on an uninitialized pointer on a certain condition.
>
> To reproduce:
>
> mkdir -p /tmp/test /tmp/plop
> openrsync -rx /tmp/test/ /tmp/plop/
>
> Result:
>
> openrsync(3470) in free(): bogus pointer (double free?) 0x7f7dcdc8
>
Frederic Cambus:
> Here is a diff to sync kernel fonts with the latest released Spleen
> version, bringing the following improvements:
[...]
> - Add some artefacts on each side of the lower case 'i' characters
Yes, please!
Actually, I still think the 'i' should have a wide foot like the '1'...
bu
On 2019-06-02, Hiltjo Posthuma wrote:
> I noticed when using openrsync with a remote and a ssh_prog set (-e option)
> the
> socket is closed twice also.
Yes, we need to decide whether rsync_socket() should close the
socket itself or not. Since rsync_socket() doesn't open the socket,
I guess it
Brian Callahan:
> lang/ponyc uses $(shell getconf LONG_BIT) in its Makefile to
> determine whether or not we're on a 64-bit platform.
> However, our getconf(1) doesn't support reporting LONG_BIT.
> This diff enables it. GNU/FreeBSD/DragonFly/MacOS getconf reports
> LONG_BIT, but NetBSD getconf doe
Ingo Schwarze:
> > Not sure what do about "ELF(3)" in elf(3).
> Index: elf.3
> Index: elf_flagdata.3
> Index: elf_open.3
> Index: gelf.3
> Index: gelf_newehdr.3
> Index: gelf_newphdr.3
-snip-
OK, but note that there are also three instances of
.Xr ELF 3
in elf.3.
--
Christian "naddy" Weisgerbe
On 2019-05-25, Artturi Alm wrote:
> doesn't look pretty, and is more lines, but in browsers the lack of .Xr
> does really stand out, and not in a good way, in my opinion. :]
Thanks, committed.
I grepped the man pages for other instances of this and found and
fixed a few. Not sure what do about
[Time to take this for ok-shopping.]
scp's ability to copy between two remote hosts has always seemed of
very limited use to me, since the first remote host must be able to
establish an ssh connection to the second host. This is really a case
where agent forwarding comes in handy. Currently you
[Time to take this for ok-shopping.]
ssh_config(5) and sshd_config(5) already allow adding '+' and
removing '-' an algorithm from the default list. Oddly, I mostly
find myself wanting to prefer an algorithm, i.e., place it at the
head of the list without removing anything. The patch below adds
t
Tweak sysupgrade's verbosity:
* If we are verifying old sets, say so. This is not an instaneous
operation.
* Only say that we are verifying the (newly downloaded) sets, if
we're actually doing so.
OK?
# rm /home/_sysupgrade/x*.tgz
# sysupgrade -knf
SHA256.sig 100% |***
Add a -k flag to sysupgrade to keep the files in /home/_sysupgrade,
since they will be deleted after the upgrade by default.
Florian already added the installer part yesterday.
OK?
Index: sysupgrade.8
===
RCS file: /cvs/src/usr.sbin/
On 2019-05-09, Christian Weisgerber wrote:
>> Does unpriv do anything if you don't give it a command to run?
>
> It should return an error code and abort the script.
And actually that's what it does.
(Sorry, I confused myself there for a moment.)
--
Christ
Currently we verify all sets, reboot into bsd.upgrade, and there the
installer re-verifies them all over. The diff below eliminates this
redundant check.
In the installer, verification is triggered by the presence of
SHA256.sig. Later, SHA256.sig is saved, so sysupgrade can exit
early if no newe
On 2019-05-08, Jarkko Oranen wrote:
>> -unpriv -f $f ftp -Vmo ${f} ${URL}${f}
>> +unpriv -f $f
>
> Does unpriv do anything if you don't give it a command to run?
It should return an error code and abort the script. I need to
investigate why that doesn't happen.
--
Christian "naddy" We
Mark Kettenis:
> The UEFI standard defines an (optional) protocol for getting random
> numbers. A while back I wrote support for this for the arm64
> bootloader. But we should use this on amd64 as well. I tested that
> this actually works on a machine with a recent AMD Ryzen APU.
This also wor
On 2019-05-03, Christian Weisgerber wrote:
> * For "unpriv -f file", chown file back to root once the command
> finishes, like install.sub does.
>
> I think it's desirable that these behave in the same way. It also
> means that we don't have to worry that
* For "unpriv -f file", chown file back to root once the command
finishes, like install.sub does.
I think it's desirable that these behave in the same way. It also
means that we don't have to worry that later unpriv commands could
have modified the file. This has consequences:
* Only verify t
It is not an error condition if no new snapshot is available.
(Is it?)
Index: sysupgrade.sh
===
RCS file: /cvs/src/usr.sbin/sysupgrade/sysupgrade.sh,v
retrieving revision 1.9
diff -u -p -r1.9 sysupgrade.sh
--- sysupgrade.sh 3 Ma
On 2019-04-27, Marco Bonetti wrote:
> +unpriv -f SHA256.sig.tmp ftp -Vmo SHA256.sig.tmp ${URL}SHA256.sig
> +TMP_SHA=$(sha256 -q SHA256.sig.tmp)
> +
> +unpriv touch SHA256.sig
This fails if SHA256.sig doesn't exist yet. The unprivileged user
cannot create files in $SETSDIR.
> +unpriv cat SHA256
If you boot efiboot via PXE and at the prompt enter a different boot
device, e.g.
boot> boot sd0a:/bsd
then efiboot will pretend to honor this
booting sd0a:/bsd: ...
but in fact ignore the device and boot from TFTP instead.
In the MI part of the boot loader, open() calls devopen(), which
Ingo Schwarze:
> here are serveral bugfixes and improvements for the rsync(1) manual.
>
> OK?
ok
I have some comments, but I think you should commit this. Further
tweaks can follow later.
> Missing information:
> * Change the misleading argument placeholder "portnumber" to "service";
>eve
Some sort of merge error; FreeBSD, where this was imported from, always
had "else if". Also, semantically more plausible.
ok?
Index: sys/dev/pci/if_wb.c
===
RCS file: /cvs/src/sys/dev/pci/if_wb.c,v
retrieving revision 1.69
diff -u -
I'm grepping the tree for "} if" lines...
I'm confident that these were intended as "else if", compare the
corresponding ror_* functions. Also, it doesn't actually change
the result.
ok?
Index: sys/dev/x86emu/x86emu.c
===
RCS file:
GPL rsync allows tunneling an rsync:// connection to a daemon through
an explicitly given -e command. Specifically, you can rsync the
repository from sthen@'s mirror through ssh this way:
$ rsync -av --del -e 'ssh -W localhost:rsync -lanoncvs' \
rsync://anoncvs.spacehopper.org/OpenBSD-CVS/ /c
On 2019-03-30, Jason McIntyre wrote:
> we could use: [--del[ete]]
> that looks ugly, but is short and clear. kind of vi(1) style.
>
> we could just note in the description of --delete that it can be
> abbreviated to --del. that would add the least amount of text, but not
> show up in usage/synops
Oops, I guess I should document --del.
Index: main.c
===
RCS file: /cvs/src/usr.bin/rsync/main.c,v
retrieving revision 1.37
diff -u -p -r1.37 main.c
--- main.c 30 Mar 2019 07:24:42 - 1.37
+++ main.c 30 Mar 2019 08:1
* Add --del as alias for --delete, because I've grown used to this
shorter option. For GPL rsync, these behave identically when
talking to rsync 3.0.0 or newer.
* Drop --no-delete. In GPL rsync, the "no-" prefix is only available
for options that are implied by other options. (Weirdly, th
Minor man page tweaks to reflect the fact that PKCS#11 support is
no longer limited to RSA.
Index: ssh-keygen.1
===
RCS file: /cvs/src/usr.bin/ssh/ssh-keygen.1,v
retrieving revision 1.156
diff -u -p -r1.156 ssh-keygen.1
--- ssh-keygen
On 2019-01-30, Martin Pieuchot wrote:
> Newer version with fewer atomics and some barrier corrections as pointed
> out by visa@.
This has now gone through a full amd64 package build without problems.
--
Christian "naddy" Weisgerber na...@mips.inka.de
The tail end of make release on arm64 looks like this:
...
===> arm64/miniroot
cp miniroot64.fs /home/rel-arm64
su build -c 'exec make sha'
cd /home/rel-arm64; date -u "+Build date: %s - %+" > BUILDINFO; cksum -a
sha256 BUILDINFO INSTALL.`arch -ks` bsd bsd.mp bsd.rd BOOTAA64.EFI
miniroot64.fs
Match OpenSSL 1.1's openssl(1) and switch the default algorithm for
the dgst command from MD5 to SHA256.
This is in line with the MD5 -> SHA256 change in the default key
derivation for the enc command, and general good crypto principles.
ok?
Index: dgst.c
Add -iter and -pbkdf2 to the usage synopsis.
Reorder option descriptions so -iter and -pbkdf2 show up alphabetically.
Add missing argument name for -iter.
ok?
Index: enc.c
===
RCS file: /cvs/src/usr.bin/openssl/enc.c,v
retrieving rev
When tweaking the protection flags for each architecture, we want
to check for the CPU architecture (think "powerpc") and not the
system architecture (think "macppc").
ok?
Index: Makefile.inc
===
RCS file: /cvs/src/gnu/usr.bin/clang/
On 2019-01-07, "Todd C. Miller" wrote:
> The main difference with -S is that you end up with two copies of
> the target file on disk for a short period of time.
Oooh, that short period can be the time between disk flushes (30s?)
with softupdates. I think that's one of the scenarios where you
ca
On 2018-12-31, Alexandre Ratchov wrote:
> Our current USB sub-system has limitations that currently allow only
> the following combinations:
> - USB v2 devices on ehci(4) only
> - USB v1 devices on uhci(4), ohci(4) or ehci(4) root hub only
>
> If you have an audio device that is class compliant
Mark Kettenis:
> I wonder though whether we can simply replace ENTRY(start_phys) with
> ENTRY(start) like we do on amd64. The bootloader strips the high bits
> from the entry point address already, so using the virtual address as
> the entry point address should work.
/me builds & boots kernel .
Correct linker emulation name for lld when building i386 kernels
on amd64.
(Does anybody build i386 kernels on amd64? Apparently nobody has
since the lld switch.)
Index: sys/arch/i386/conf/Makefile.i386
===
RCS file: /cvs/src/sys/ar
Move the calculation of the entry point to after the description
of the .text section. LOADADDR(.text) is not available earlier
with lld.
With this, lld builds working kernels on i386.
Also: Should this be __start_phys? It seems odd that we put a
non-reserved symbol into the kernel namespace.
I ran across this:
$ readelf -h /usr/local/bin/w3m
...
readelf(71968) in free(): bogus pointer (double free?) 0x1
Abort trap (core dumped)
In readelf.c there's a static arrary:
static bfd_vma dynamic_info[DT_JMPREL + 1];
Later this array is written to like this:
switch (entry->d_ta
Mark Kettenis:
> > Currently, building an amd64 or i386 kernel with lld breaks in
> > makegap.sh. It turns out that we already have the missing pieces
> > (gapdummy) on some other archs, so we can just copy them.
>
> Any reason why you didn't commit this yet?
Theo was concerned that this might
Provide an MD 64-bit byteswapping function build on 32-bit swaps
as we do on arm and i386. Copied from arm.
If there are no MD byteswapping functions, MI macros are used.
These are wrapped by static inline functions to prevent multiple
evaluation of their argument. If there are MD functions, the
The sys/arch/*/include/endian.h headers were the only place in the
whole source tree that used __statement(), and they did so to mark
statement expressions, which are a GNU extension.
With this single consumer gone, we can garbage-collect the __statement()
macro itself. I don't think this usage w
Currently, building an amd64 or i386 kernel with lld breaks in
makegap.sh. It turns out that we already have the missing pieces
(gapdummy) on some other archs, so we can just copy them.
With this, I can successfully build and run
* amd64 kernel with ld.bfd
* i386 kernel with ld.bfd
* amd64 kerne
Some architectures generate a gapdummy.c file for kernel builds.
Add this to the "clean" target.
OK?
Index: arch/arm64/conf/Makefile.arm64
===
RCS file: /cvs/src/sys/arch/arm64/conf/Makefile.arm64,v
retrieving revision 1.28
diff -u -
There are some instances under src/gnu of Makefile.bsd-wrapper
failing to pass variables like CC to a configure script. As a
result, the configure tests are actually run with gcc (!) on amd64
and i386. The patch below fixes
lib/libiberty
usr.bin/binutils
usr.bin/binutils-2.17
usr.bin/cvs
In the pkill man page, clarify that the patterns are EXTENDED regular
expressions. I shouldn't have to guess, try, or look at the source
for this.
ok?
Index: pkill.1
===
RCS file: /cvs/src/usr.bin/pkill/pkill.1,v
retrieving revision
Is there a reason ssh doesn't consult services(5) for port numbers?
This has irked me forever. I'd rather write ssh -L icb:localhost:icb
instead of ssh -L 7326, wait, 7236, uhm, grep icb /etc/services...
I don't think there is any syntactic ambiguity since Unix sockets
already must contain a '/'
I noticed this curious error in a package bulk build:
Use of uninitialized value $pid in numeric eq (==) at
/usr/libdata/perl5/LT/Exec.pm line 96.
Perl's fork() wrapper returns undef rather than -1 on error.
Handle this correctly. OK?
Index: LT/Exec.pm
==
Since the VAX port was discontinued, all our remaining architectures
use IEEE floating point, as will any future ones.
For this reason, I would like to remove all of libm/noieee_src,
which is unused now. I don't quite know how to actually prove that
it is unused, but
* I grepped the full src tree
Marco Pfatschbacher:
> I noticed that ls(1) defaults to multi-column output,
> even if the output is piped to another command.
I guess this is intentional?
> ls(1) can detect whether stdout isatty(3) all by itself.
> I think it's time to remove this artifact from 1996.
Even the oldest version o
Mark Kettenis:
> As we discussed some time ago, adding symlinks like
>
> libfoo.so -> libfoo.so.x.y
>
> brings us more in line with other ELF platforms and will allow us to
> drop some OpenBSD-specific code from our linkers. It also makes it
> possible for developers to easily revert a librar
On 2018-04-20, Mark Kettenis wrote:
> At some point I tried playing with this on the APU, which has the v3
> engine I think. Couldn't get anything to work (but I didn't try very
> hard). Would be interesting to see whether this actually works on
> that machine.
It attaches. How can I tell if
David Hill:
> This diff adds sizes to free(), which completes ufs/ffs.
It's broken at least for softdep+UFS2. This chunk blows up:
> --- ufs/ffs/ffs_softdep.c 10 Feb 2018 05:24:23 - 1.138
> +++ ufs/ffs/ffs_softdep.c 29 Mar 2018 02:55:37 -
> @@ -4034,7 +4036,8 @@ handle_writ
Fix a compiler warning about a missing prototype for tolower() when
building the installer's cut down sysctl.
ok?
Index: distrib/special/sysctl/sysctl.c
===
RCS file: /cvs/src/distrib/special/sysctl/sysctl.c,v
retrieving revision 1.1
On 2018-02-07, Claudio Jeker wrote:
> On Wed, Feb 07, 2018 at 07:28:42PM +0100, Ingo Schwarze wrote:
>> while trying to fix the documentation of screen splitting in the vi(1)
>> manual page, jmc@ suggested to stop documenting it at all. I like
>> the idea because splitting is useless and confusi
On amd64 and i386, when we recognize that we've been PXE-booted and
identified the interface, there is this:
...
if (ifp) {
#if defined(NFSCLIENT)
printf("PXE boot MAC address %s, interface %s\n",
ether_sprintf(bios_bootmac->mac),
On 2018-02-04, Patrick Wildt wrote:
>> this diff allows us to load the Intel microcode much earlier.
I'm trying to understand the twisty logic here.
There are three cases:
(1) old rev < update rev
=> update gets applied
(2) old rev == update rev
=> cpu_ucode_intel_match() returns succe
Alexei Malinin:
> If the base system supported multi-booting in MBR would the community be
> interested in it?
It should be easy to adapt FreeBSD's boot0. I don't remember any
demand for it, and now MBR is on its way out...
https://svnweb.freebsd.org/base/head/stand/i386/boot0/boot0.S
boot0 +
This allows an arm64 kernel to recognize that it has been netbooted
and to add the boot interface to the "netboot" group. efiboot grabs
the MAC address from the PXE environment, passes it to the kernel,
where it is matched against the list of ethernet interfaces and the
boot device is set. Concep
The nam2blk[] array contains MD mappings between block device names and
major device numbers.
This patch syncs the nam2blk entries with the bdevsw table, which
is the definitive list of block devices supported on an architecture.
This fixes the non-OpenBSD entries on arm64, removes obsolete
mentio
Fix TFTP reading of zero-size files in efiboot:
The AllocatePages EFI call returns an error when the allocation
size is 0. Skip allocating memory and actually transferring the
file when it is empty.
Properly return the number of unread bytes so that a read() of n
bytes does not return n if no by
tftpd needs to handle the transfer size option when faking up
/etc/random.seed.
Without this, clients that query the size receive 0 as response,
preventing efiboot from loading random.seed.
ok?
Index: tftpd.c
===
RCS file: /cvs/src/
Daniel Loebenberger:
> - The construction of SHA3 differs considerably from the SHA2
> constructions
> - SHA3's design principles are far better understood than the ones of
> SHA2.
I hear you, but you are missing the point.
> - A possible migration away from SHA2 will be
> faster when including
On 2018-01-09, Daniel Loebenberger wrote:
> enclosed you find a patch to add basic SHA3-/Keccak support to OpenBSD.
What do you want to use it for?
I've had something like this on my to-do list for some time and
eventually dropped it because I couldn't think of a compelling use
case.
"Maybe Sk
On 2017-11-17, Stefan Sperling wrote:
> This diff makes the WPA key available only if the interface is in
> debug mode (suggestion by phessler). If this is acceptable then I
> can also try to squeeze a hint into the ifconfig man page so that
> this mechanism can be discovered by those who don't r
On 2017-11-17, David Gwynne wrote:
> can we have modified displays within a view?
We already have this for the ifstat view.
The character B changes the counter view between bytes and
bits. Pressing b displays statistics as calculated from boot
time. r changes the co
Martin Pieuchot:
> Diff below makes them use rtisvalid(9) instead which check for the
> RTF_UP flag. This should make many "route contains no arp information"
> messages disappear.
>
> But I'm afraid this can expose another weird driver/NFS setup were non
> UP routes are used,
Indeed.
Booting
On 2017-11-01, Jan Stary wrote:
> Why does awk need to fiddle with LC_NUMERIC?
You don't want to parse your awk program in a locale where the
decimal separator is a comma.
--
Christian "naddy" Weisgerber na...@mips.inka.de
As far as I can tell, nothing in the build system generates .depend
files any longer, nor do they appear as prerequisites in any make
rules. The patch below removes all remaining mentions of .depend
(lists of files to clean, spurious .depend targets, etc) from the
src Makefiles.
This passed a mak
Christian Weisgerber:
> Admittedly, this is cosmetic:
> Use the modern POSIX idiom "-exec ... {} +" instead of find|xargs.
tb@ kindly pointed out to me that Klemens Nanni submitted a better
patch for this four months ago; "-exec +" allows us to combine the
two find i
Admittedly, this is cosmetic:
Use the modern POSIX idiom "-exec ... {} +" instead of find|xargs.
ok?
Index: include/Makefile
===
RCS file: /cvs/src/include/Makefile,v
retrieving revision 1.219
diff -u -p -r1.219 Makefile
--- include/M
Scott Cheloha:
> Use a monotonic clock for the elapsed time trial.
FreeBSD uses getrusage() to fetch the user time used. I think that
makes more sense.
https://svnweb.freebsd.org/base/head/sbin/md5/md5.c?revision=307658&view=markup#l293
--
Christian "naddy" Weisgerber
On 2017-09-07, Christian Weisgerber wrote:
> However, I don't know how to test that these functions actually work.
FWIW, I checked all *.o files from a make build of base, but there
are no references to __float(di|undi)[dsx]f. I guess nothing uses
these functions and clang doesn&#x
Christian Weisgerber:
> > Maybe this would already help? Would at least not throw stones into the
> > way of the next person doing an upgrade of compiler-rt...
> -snip-
>
> Yes, that's better... but errors out:
> make: don't know how to make floatdixf.c.c (p
Patrick Wildt:
> Maybe this would already help? Would at least not throw stones into the
> way of the next person doing an upgrade of compiler-rt...
-snip-
Yes, that's better... but errors out:
make: don't know how to make floatdixf.c.c (prerequisite of: floatdixf.c.o)
--
Christian "naddy" We
Somebody noticed this on FreeBSD:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221733
On i386, libcompiler_rt includes assembly implementations for
floating point conversion functions that unconditionally use SSE2
instructions, which are not supported by older CPUs. Specifically,
these files
Edd Barrett:
> > lang/pypy retguard
>
> I'm looking into PyPy.
>
> Can you provide the build output?
Oops, sorry, I forgot to save the logs before I cleaned the machines
yesterday. Should be straightforward to reproduce, though. At
some point the build dies with a segfault.
On 2017-08-19, Theo de Raadt wrote:
> We are at the point where userland and base are fully working without
> regressions, and the remaining impacts are in a few larger ports which
> directly access the return address (for a variety of reasons).
Fallout from building amd64 packages with a retgua
Jonathan Matthew:
> Better version that actually preserves the port command register state across
> resets, rather than throwing it away and replacing it with garbage:
This appears to break ahci on the OverDrive 1000:
...
ahci0 at simplebus0: AHCI 1.3
scsibus0 at ahci0: 32 targets
ahci0: stoppin
Martin Pieuchot:
> The second one has been found by Matthias Pitzl and fixed this week
> by bluhm@: em(4) was reporting incorrect link state. This link state
> being applied to the corresponding routes, rtisvalid() would refuse
> to use the selected route which would break some NFS setups.
>
> P
Mark Kettenis:
> The diff below fixes the
>
> /usr/bin/ld: /tmp/BC-10602a.o: relocation R_X86_64_PC32 against `xxx' can not
> be used when making a shared object; recompile with -fPIC
>
> errors that we see with some ports when compiled with clang.
That's great news!
> should go through a fu
David Gwynne:
> secondly, im always wary of truncating hash output in case it throws away
> some of the guarantees it's supposed to provide. if you cut sha512 output
> down to an 8th of its size, is it 8 times easier to calculate a collision, or
> more than 8 times easier? sha384 being a trunc
On 2017-07-14, Florian Obser wrote:
> --- etc/rc
> +++ etc/rc
> @@ -47,6 +47,14 @@ update_limit() {
> done
> }
>
> +# Apply soii.key settings.
> +soii_key() {
> + stripcom /etc/soii.key |
> + while read _line; do
> + sysctl -q "net.inet6.ip6.soiikey=$_line"
> + do
On 2017-07-13, Florian Obser wrote:
> It switches the hash function to SipHash24 from sha512 as suggested by dlg
It's for from clear to me whether SipHash is suitable for crypto
operations, and which ones, other than the hash tables it was
designed for.
We went with SHA-512 for RFC 1948 TCP ini
On 2017-07-11, Florian Obser wrote:
> + SHA512Final(digest, &ctx);
> +
> + /* assumes sizeof(digest) > sizeof(ifid) */
> + bcopy(digest, &in6->s6_addr[8], 8);
... and we're finished here.
> + /* make sure to set "u" bit to local, and "g" bit to individual. */
> + in6->s6_add
On 2017-06-29, "Anthony J. Bentley" wrote:
> The vi manual is long enough. It doesn't seem very helpful to keep stubs
> that have been stubs for 20 years.
Should we retain some documentation along the lines of "nex/nvi
does not implement the following options from the original ex/vi:
..."?
--
Stefan Sperling:
> inet 0.0.0.0 255.255.255.255 NONE \
> pppoedev em0 authproto pap \
> authname 'testcaller' authkey 'donttell' \
>dynaddr dyndest up
>
> Tested on my ADSL line and it works as expected.
The example in the pppoe(4) m
Fix clang warning about tautological compare: an unsigned long can't
be negative.
ok? Or dlg@ can slip it in with his next ifconfig change.
Index: brconfig.c
===
RCS file: /cvs/src/sbin/ifconfig/brconfig.c,v
retrieving revision 1.14
Todd C. Miller:
> I think you want 0xU, not 0xL. Otherwise you will
> have the same issue on i386.
???
We need a constant that comes out as the "long" value
0x on 64-bit platforms
0x on 32-bit platforms (degenerate case)
--
Christian "
This catches up i386 with the changes to acpi_wakecode.S and mptramp.s
that were made on amd64 for clang. Here's kettenis@'s original
commit message:
--->
Generating mixed 16-bit/32-bit/64-bit code with clang's integrated
assembler is a bit tricky. It supports the .code16, .code3
201 - 300 of 589 matches
Mail list logo
