gilles [Wed, 14 Oct 2015 21:30:40 +0000 (21:30 +0000)]
SMTPD_MAXPATHLEN -> PATH_MAX, this was unnoticed as file is not linked
gilles [Wed, 14 Oct 2015 21:27:29 +0000 (21:27 +0000)]
imsg_read() may return EAGAIN, handle it in mproc_dispatch()
beck [Wed, 14 Oct 2015 21:25:16 +0000 (21:25 +0000)]
Bail out early if we have no buf_len
ok miod@
gilles [Wed, 14 Oct 2015 21:18:19 +0000 (21:18 +0000)]
whitespace + only log TRACE_MPROC if not IMSG_STAT_{IN,DE}CREMENT
tedu [Wed, 14 Oct 2015 21:12:10 +0000 (21:12 +0000)]
fix a memory leak reported by Qualys Security.
move the bndec variable in tighter since it's not used elsewhere in the
loop, then always free it after use.
ok bcook miod
gilles [Wed, 14 Oct 2015 21:05:31 +0000 (21:05 +0000)]
remove a handful of log_warn that we should handle at a different place to
make them really useful
gilles [Wed, 14 Oct 2015 21:04:37 +0000 (21:04 +0000)]
whitespace
gilles [Wed, 14 Oct 2015 21:02:11 +0000 (21:02 +0000)]
fix define and enhanced status code reason for 5.5.0
beck [Wed, 14 Oct 2015 21:02:08 +0000 (21:02 +0000)]
Ensure we don't write a 0 byte past end of the buffer in the error case.
ok bcook@ deraadt@
deraadt [Wed, 14 Oct 2015 20:57:28 +0000 (20:57 +0000)]
Use a strict $PATH of "/usr/bin:/usr/local/bin" to run the (de)compressors
(gzip, compress, bzip2) rather than following the user's path. This
seems easier than hardcoding the paths elsewhere and using basename().
pax/tar is pledged itself, but it can spawn one of these programs if
asked. The three found at the strict path use pledge "stdio" very early
during startup, providing a warm fuzzy pledge->exec->no-pledge->pledge
interlock. For bzip2, this assumes use of the ports/packages version
installed to /usr/local/bin, which has been pledged by sthen@.
Doing a 'tar tvfz hostile.tgz' becomes a bit safer, since an attacker
finding a buffer overflow or use after free has significantly fewer
system calls available (only pledge "stdio" in the decompressor).
ok millert sthen
gilles [Wed, 14 Oct 2015 20:57:17 +0000 (20:57 +0000)]
we dump esc_code if we have an esc_class, code may be (and actually was)
confused as a pasto, so add a comment to make it clear that this is not
an error
millert [Wed, 14 Oct 2015 20:54:07 +0000 (20:54 +0000)]
Allow group wheel to read the mail log. OK gilles@ sthen@
gilles [Wed, 14 Oct 2015 20:51:53 +0000 (20:51 +0000)]
whitespace
gilles [Wed, 14 Oct 2015 20:45:30 +0000 (20:45 +0000)]
whitespace
gilles [Wed, 14 Oct 2015 19:56:58 +0000 (19:56 +0000)]
pledge() privileged process
ok deraadt@
gilles [Wed, 14 Oct 2015 19:39:16 +0000 (19:39 +0000)]
pledge() pony and lookup
ok deraadt@
deraadt [Wed, 14 Oct 2015 19:14:32 +0000 (19:14 +0000)]
sync
deraadt [Wed, 14 Oct 2015 17:54:46 +0000 (17:54 +0000)]
worse, need even more includes...
tobias [Wed, 14 Oct 2015 17:29:44 +0000 (17:29 +0000)]
Check mmap and read return values. While at it, remove unused duplicated file.
ok millert@
florian [Wed, 14 Oct 2015 17:26:01 +0000 (17:26 +0000)]
To specify a source address ping uses -I while ping6 uses -S. Switch
ping6 -I to the ping-alike semantics.
sthen@ thinks this is OK
deraadt [Wed, 14 Oct 2015 17:06:58 +0000 (17:06 +0000)]
pledge "stdio rpath" is good enough for these mainline BSD auth login
programs.
(I am very surprised pledge ended up working for programs like this)
ok semarie millert
deraadt [Wed, 14 Oct 2015 16:58:55 +0000 (16:58 +0000)]
Since the fsck_* programs now only handle one filesystem, this creates
a point where open() and disklabel reading have completed. After that
point, pledge "stdio".
As a result, an fsck of a hostile partition (noone ever does that, or
do they? :) is done by a program with SUBSTANTIALLY less system call
exposure.
ok semarie
tobias [Wed, 14 Oct 2015 16:42:51 +0000 (16:42 +0000)]
Copy permissions AND ownership when -o will override an input file.
with input by and ok deraadt@, millert@, tim@
deraadt [Wed, 14 Oct 2015 16:29:37 +0000 (16:29 +0000)]
sync
deraadt [Wed, 14 Oct 2015 15:54:49 +0000 (15:54 +0000)]
unfortunately rewritelabel() just before termination does a non-permitted
ioctl to rewrite the label, in support of the old-school "frag info in
the disklabel" concept. disklabel folk, please come talk to me...
reyk [Wed, 14 Oct 2015 14:51:57 +0000 (14:51 +0000)]
sync with httpd - no functional change, just C99 types
deraadt [Wed, 14 Oct 2015 14:33:45 +0000 (14:33 +0000)]
Only accept one filesystem/device as argument for checking. Few people
will be calling these directly, and not for the multiple filesystem case.
fsck(8) is generally the parent and will handle things.
ok semarie; this change will also help a goal jsing has
deraadt [Wed, 14 Oct 2015 14:24:03 +0000 (14:24 +0000)]
When pledged with "fattr", allow chown to supplimentary groups. This
came out of a discussion regarding "sort foo -o foo".
ok semarie
jsg [Wed, 14 Oct 2015 14:13:12 +0000 (14:13 +0000)]
add includes for crc32() and uuid_dec_be() missed in rev 1.11
ok krw@
jsg [Wed, 14 Oct 2015 13:59:31 +0000 (13:59 +0000)]
Init a variable in the recently added carp_vhe_match() function clang
and mpi believe could be used uninitialised.
ok mpi@
espie [Wed, 14 Oct 2015 13:52:11 +0000 (13:52 +0000)]
gc lst_ForEachNodeWhile, which isn't actually in use anywhere
espie [Wed, 14 Oct 2015 13:50:22 +0000 (13:50 +0000)]
make sure we use stdbool.h
Mostly diff by Daniel Dickman, who told me to commit in his stead,
as he's tied up at work.
jsg [Wed, 14 Oct 2015 13:32:44 +0000 (13:32 +0000)]
include err.h for the err() calls added in rev 1.46
jsg [Wed, 14 Oct 2015 13:27:50 +0000 (13:27 +0000)]
Don't use the NONE enum value where NULL was intended. Found with clang.
ok renato@
vgross [Wed, 14 Oct 2015 13:23:25 +0000 (13:23 +0000)]
bugfix: add ${.CURDIR} to deal with obj symlinks
semarie [Wed, 14 Oct 2015 13:22:56 +0000 (13:22 +0000)]
unbreak regress/sbin/newfs
/dev/prandom is no more since Nov 30, 2008
OK otto@
vgross [Wed, 14 Oct 2015 13:17:33 +0000 (13:17 +0000)]
bugfix : use ${.CURDIR} to cope with obj symlinks
vgross [Wed, 14 Oct 2015 12:47:07 +0000 (12:47 +0000)]
add regress tests for automatic port allocation.
- enable ipv4
- leave ipv6 disabled
vgross [Wed, 14 Oct 2015 12:38:52 +0000 (12:38 +0000)]
add regress tests for automatic port allocation
mpi [Wed, 14 Oct 2015 10:18:03 +0000 (10:18 +0000)]
Reset the RTF_CONNECTED flag when cloning an entry.
While here check for RTF_CLONED insted of RTM_RESOLVE when adding an
entry.
Found while debugging naddy@'s NFS vs em(4) vs rtisvalid(9) issue.
mpi [Wed, 14 Oct 2015 10:09:30 +0000 (10:09 +0000)]
Rewrite the logic around the dymanic array of routing tables to help
turning rtable_get(9) MP-safe.
Use only one per-AF array, as suggested by claudio@, pointing to an
array of pointers to the routing table heads.
Routing tables are now allocated/initialized per-AF. This will let
us allocate routing table on-demand instead of always having an
AF_INET, AF_MPLS and AF_INET table as soon as a new rtableID is used.
This also get rid of the "void ***" madness.
ok dlg@, jmatthew@
sunil [Wed, 14 Oct 2015 09:14:11 +0000 (09:14 +0000)]
Convert fgetln to getline.
Ok millert@ eric@ gilles@
schwarze [Wed, 14 Oct 2015 09:11:25 +0000 (09:11 +0000)]
tweak previous (two details i apparently missed)
doug [Wed, 14 Oct 2015 08:12:12 +0000 (08:12 +0000)]
Pledge "stdio" for simple games.
ok semarie@
reyk [Wed, 14 Oct 2015 08:02:38 +0000 (08:02 +0000)]
Two more char -> unsigned char in ctype functions.
reyk [Wed, 14 Oct 2015 07:58:14 +0000 (07:58 +0000)]
More (unsigned char) casts for ctype functions.
Pointed out by Michael McConville
reyk [Wed, 14 Oct 2015 07:41:28 +0000 (07:41 +0000)]
Add EVP_AEAD_CTX_init(3) manpage to document the new(ish) AEAD API.
The "authenticated encryption with additional data" API is used for
ciphers like AES-GCM or ChaCha20-Poly1305. The manpage is a beginning
and certainly needs more work, especially improvements in the EXAMPLES
section.
Based on agl's source code comments.
Converted from pod to mandoc by schwarze@
OK schwarze@ jsing@
semarie [Wed, 14 Oct 2015 07:19:23 +0000 (07:19 +0000)]
enable pledge(2) in rain(6)
it is libcurses program: at init it needs "stdio rpath getpw tty", and after
drop to just "stdio tty". "tty" is needed at end for restoring the tty.
initial patch from doug@
ok doug@ deraadt@
guenther [Wed, 14 Oct 2015 04:55:17 +0000 (04:55 +0000)]
Remove conditional compilation and #defines around signal handling
Don't catch signals that were ignored on entry
Suppress SIGCHLD if our kid is stopped: we don't care and it's not an error
ok millert@
deraadt [Wed, 14 Oct 2015 04:05:43 +0000 (04:05 +0000)]
pledge "tty" can allow ioctl TIOCEXCL on a tty
deraadt [Wed, 14 Oct 2015 04:03:01 +0000 (04:03 +0000)]
I messed up reading the call graph. -d delete does use search, so a
late pledge is not possible in this way.
deraadt [Wed, 14 Oct 2015 03:27:02 +0000 (03:27 +0000)]
sendmsg() is allowed to pass cmsg's which are not CMSG_RIGHTS - last
refactoring inverted the checks; spotted by sthen in ping6.
krw [Wed, 14 Oct 2015 00:19:04 +0000 (00:19 +0000)]
Backout last. Breaks sparc64, amoung other dubiousness.
requested by deraadt@
schwarze [Tue, 13 Oct 2015 23:30:42 +0000 (23:30 +0000)]
Reject the escape sequences \[uD800] to \[uDFFF] in the parser.
These surrogates are not valid Unicode codepoints,
so treat them just like any other undefined character escapes:
Warn about them and do not produce output.
Issue noticed while talking to stsp@, semarie@, and bentley@.
schwarze [Tue, 13 Oct 2015 22:57:49 +0000 (22:57 +0000)]
Major character table cleanup:
* Use ohash(3) rather than a hand-rolled hash table.
* Make the character table static in the chars.c module:
There is no need to pass a pointer around, we most certainly
never want to use two different character tables concurrently.
* No need to keep the characters in a separate file chars.in;
that merely encourages downstream porters to mess with them.
* Sort the characters to agree with the mandoc_chars(7) manual page.
* Specify Unicode codepoints in hex, not decimal (that's the detail
that originally triggered this patch).
No functional change, minus 100 LOC, and i don't see a performance change.
tim [Tue, 13 Oct 2015 22:14:51 +0000 (22:14 +0000)]
Call the sort program through $SORTPROG, as intended; OK millert@
mpi [Tue, 13 Oct 2015 20:57:46 +0000 (20:57 +0000)]
Prevent a NULL-pointer dereference when closing a ugen(4) node
in case the kernel failed to change the interface of a device.
Found the hardway by okan
millert [Tue, 13 Oct 2015 20:55:37 +0000 (20:55 +0000)]
In rev 1.15 the sizeof argument was fixed in a strlcat() call but
the truncation check immediately following it was not updated to
match. Not an issue in practice since the buffers are the same
size. OK deraadt@
lum [Tue, 13 Oct 2015 20:10:09 +0000 (20:10 +0000)]
Check if a file name can be extracted from a line before marking for
deletion.
deraadt [Tue, 13 Oct 2015 20:00:49 +0000 (20:00 +0000)]
3 more headers required for one stinking inet6 ioctl..
sashan [Tue, 13 Oct 2015 19:32:31 +0000 (19:32 +0000)]
- pf_insert_src_node(): global argument (arg6) is useless, function
always gets pointer to rule.
- pf_remove_src_node(): function should always remove matching src node,
regardless the sn->rule.ptr being NULL or valid rule
- sn->rule.ptr is never NULL, spotted by mpi and Richard Procter _von_ gmail.com
OK mpi@, OK mikeb@
ajacoutot [Tue, 13 Oct 2015 18:52:56 +0000 (18:52 +0000)]
sm_error() already does the exit for us.
deraadt [Tue, 13 Oct 2015 18:42:06 +0000 (18:42 +0000)]
Allow ioctl SIOCGNBRINFO_IN6 in case of "route", for use by ndp.
deraadt [Tue, 13 Oct 2015 18:38:34 +0000 (18:38 +0000)]
NET_RT_FLAGS must also accept a proto selection.
tobias [Tue, 13 Oct 2015 17:07:05 +0000 (17:07 +0000)]
Test t16 for ed-formatted diffs does not contain a substitution.
Add a minimalistic check in t17.
tim [Tue, 13 Oct 2015 16:55:03 +0000 (16:55 +0000)]
Ignore the setuid/setgid/sticky bits when copying the permissions of an input
file to the new output file. In preparation for pledge(2).
Suggested by and OK millert@
tobias [Tue, 13 Oct 2015 16:37:17 +0000 (16:37 +0000)]
Replace our /^\.\././ expression with /.//. The term is simpler and has
the same meaning in our diff ed-context.
As a bonus, our ed-diff output can be processed by GNU patch now, too.
okay millert@
jsing [Tue, 13 Oct 2015 16:31:08 +0000 (16:31 +0000)]
Put ASN1_dup() under #ifndef LIBRESSL_INTERNAL.
deraadt [Tue, 13 Oct 2015 16:30:55 +0000 (16:30 +0000)]
After the socket is open, the remainder is just io operations.
Use pledge "stdio".
florian [Tue, 13 Oct 2015 16:26:54 +0000 (16:26 +0000)]
Remove -b flag and let ping6 set the socket buffer size automatically
like ping.
Suggested by deraadt@, OK dlg
tim [Tue, 13 Oct 2015 16:21:42 +0000 (16:21 +0000)]
-C and -c allow at most one input file. Ensure this is the case when the
input files are specified through --files0-from.
OK millert@
djm [Tue, 13 Oct 2015 16:15:21 +0000 (16:15 +0000)]
apply PubkeyAcceptedKeyTypes filtering earlier, so all skipped
keys are noted before pubkey authentication starts. ok dtucker@
deraadt [Tue, 13 Oct 2015 16:09:24 +0000 (16:09 +0000)]
allow getsockopt IP_RECVDSTPORT & IPV6_RECVDSTPORT for an "inet" pledge
deraadt [Tue, 13 Oct 2015 15:55:44 +0000 (15:55 +0000)]
semarie points out i am already forgetting the rules are very tight around
*chown, even "proc fattr" won't let you do such a job. remove early pledge(),
only leave call after fchown, before when symbol table work gets done.
schwarze [Tue, 13 Oct 2015 15:50:15 +0000 (15:50 +0000)]
Reduce the amount of code by moving the three copies of the ohash
callback functions into one common place, preparing for the use of
ohash for some additional purposes. No functional change.
deraadt [Tue, 13 Oct 2015 15:43:19 +0000 (15:43 +0000)]
oops, a chown appears late on the code. to satisfy this pledge
"stdio rpath wpath cpath getpw fattr proc" early on; "proc fattr"
allows doing work with other uids on the file. after opening the
db, do the chown (replace with fchown since we know fd) and then
pledge "stdio rpath"; "rpath" due to tmpfile rename() at the end.
mistake spotted by mpi
jsing [Tue, 13 Oct 2015 15:25:18 +0000 (15:25 +0000)]
Convert ECParameters_dup() from a macro that uses ASN1_dup_of() into an
actual function. This removes the last ASN1_dup_of usage from the tree.
Feedback from doug@ and miod@
deraadt [Tue, 13 Oct 2015 15:15:30 +0000 (15:15 +0000)]
can pledge "stdio" after opening device.
deraadt [Tue, 13 Oct 2015 15:14:26 +0000 (15:14 +0000)]
pledge "stdio rpath wpath cpath fattr"; fattr due to locking code borrowed
from mail.local
deraadt [Tue, 13 Oct 2015 15:12:53 +0000 (15:12 +0000)]
pledge "stdio rpath wpath cpath", full path handling to satisfy dbopen()
deraadt [Tue, 13 Oct 2015 15:11:48 +0000 (15:11 +0000)]
pledge "stdio rpath wpath cpath proc exec". creates files throughout
it's lifetime, and often runs a pipe through "cpp"
deraadt [Tue, 13 Oct 2015 15:10:30 +0000 (15:10 +0000)]
pledge "stdio rpath wpath cpath"
stsp [Tue, 13 Oct 2015 14:36:15 +0000 (14:36 +0000)]
Don't use exp2f(), it breaks build on vax. Use a shift instead.
reported by deraadt@
jsing [Tue, 13 Oct 2015 14:03:26 +0000 (14:03 +0000)]
Convert a number of the old ASN1_{d2i,i2d}_{bio,fp}_of() macros to
ASN1_item_{d2i,i2d}_{bio,fp}() function calls.
ok beck@ doug@
jsing [Tue, 13 Oct 2015 13:59:45 +0000 (13:59 +0000)]
Single byte read/write tests.
jsing [Tue, 13 Oct 2015 13:58:33 +0000 (13:58 +0000)]
Add test coverage for peer certificate info and connection info.
jsing [Tue, 13 Oct 2015 12:56:20 +0000 (12:56 +0000)]
Make regress work again post hackathon tls_handshake/tls_read/tls_write
changes.
jsing [Tue, 13 Oct 2015 12:31:06 +0000 (12:31 +0000)]
Group d2i/i2d function prototypes by type and add missing externs for the
DSAPublicKey, DSAPrivateKey and DSAparams ASN1_ITEMs.
millert [Tue, 13 Oct 2015 12:25:04 +0000 (12:25 +0000)]
Sync printf family return value with ISO C which specifies that
these functions return a negative value on failure.
OK doug@ deraadt@
eric [Tue, 13 Oct 2015 11:32:47 +0000 (11:32 +0000)]
Add a helper for writing the message to simplify the code.
It also fixes the reported message length by taking prepended
headers into account and adds missing error checks there.
ok millert@ gilles@
gilles [Tue, 13 Oct 2015 11:03:30 +0000 (11:03 +0000)]
pledge() queue process
ok deraadt@
gilles [Tue, 13 Oct 2015 10:59:04 +0000 (10:59 +0000)]
pledge() control process
ok deraadt@
mpi [Tue, 13 Oct 2015 10:29:16 +0000 (10:29 +0000)]
Make use of rtisvalid(9) to check if local route entries match existing
configured addressses.
ok mikeb@
mpi [Tue, 13 Oct 2015 10:21:27 +0000 (10:21 +0000)]
Simplify arptfree() to no longer look at the route entry's refcounter.
ARP entries with an expired timeout are now removed from the tree even
if they are cached somehwere else. This also reduces differences with
NDP.
ok bluhm@
mpi [Tue, 13 Oct 2015 10:16:17 +0000 (10:16 +0000)]
Use rtisivalid(9) to check if the given (cached) route can be used.
Note that after calling rtalloc(9) we only check if a route has been
returned or not and do not check for its validity. This cannot be
improved without a massive refactoring.
The kernel currently *do* use !RTF_UP route due to a mismatch between
the value of ifp->if_link_state and the IFF_UP|IFF_RUNNING code.
I'd explain the RTF_UP flag as follow:
. If a cached route entry w/o RTF_UP is passed to ip{6,}_output(),
. call rtalloc(9) to see if a better entry is present in the tree.
This is enough to support MPATH and route cache invalidation.
ok bluhm@
mpi [Tue, 13 Oct 2015 09:59:37 +0000 (09:59 +0000)]
Make sure RTF_LOCAL route entries are UP when added to the tree.
This is required to maintain the original BSD behavior of locally
configured addresses being always reacheable.
Some interfaces are^w^W^Wem(4) is special and generally has a DOWN
link state when configured by netstart(8). As a result all the
route entries cloned/added before its link state goes to UP are also
marked as DOWN.
Note that this problem was not present when local addresses were
attached to lo0.
ok mikeb@
guenther [Tue, 13 Oct 2015 09:11:48 +0000 (09:11 +0000)]
Initialize va_filerev in vattr_null() to avoid leaking stack garbage;
problem pointed out by Martin Natano (natano (at) natano.net)
Also, stop chaining assignments (foo = bar = baz) in vattr_null().
The exact meaning of those depends on the order of the sizes-and-
signednesses of the lvalues, making them fragile: a statement here
mixed *six* types, but managed to get them in a safe order. Delete
a 20+ year old XXX comment that was almost certainly bemoaning a bug
from when they were in an unsafe order.
ok deraadt@ miod@
guenther [Tue, 13 Oct 2015 08:53:43 +0000 (08:53 +0000)]
ctype functions isxdigit() expect an unsigned char value; add missing casts
and adjust variable types to get correct behavior
ok beck@ millert@
guenther [Tue, 13 Oct 2015 08:49:51 +0000 (08:49 +0000)]
To alter just the atime of the mailspool, use utimensat()+UTIME_OMIT instead
of stat()+utimes(). Prefer clock_gettime() over gettimeofday() to avoid
timeval->timespec conversion
ok millert@
sunil [Tue, 13 Oct 2015 08:33:06 +0000 (08:33 +0000)]
Plug a leak.
Ok gilles@, reyk@