Commit Graph

485 Commits

Author SHA1 Message Date
christos
2964f15a7a Grr, gcc-5.4 creates DW_AT_typedef without DW_AT_type for HARD_REG_SET!
<1><c26b>: Abbrev Number: 4 (DW_TAG_typedef)
    <c26c>   DW_AT_name        : (indirect string, offset: 0x16e30): HARD_REG_ELT_TYPE
    <c270>   DW_AT_decl_file   : 57
    <c271>   DW_AT_decl_line   : 43
    <c272>   DW_AT_type        : <0x70>
 <1><c276>: Abbrev Number: 102 (DW_TAG_typedef)
    <c277>   DW_AT_name        : (indirect string, offset: 0x2f954): HARD_REG_SET
    <c27b>   DW_AT_decl_file   : 57
    <c27c>   DW_AT_decl_line   : 54
2016-06-08 21:32:27 +00:00
chs
760a7ed1e9 apply the ustack() parts of freebsd r211608:
r211608 | rpaulo | 2010-08-22 03:53:32 -0700 (Sun, 22 Aug 2010) | 8 lines

	Kernel DTrace support for:
	o uregs  (sson@)
	o ustack (sson@)
	o /dev/dtrace/helper device (needed for USDT probes)

	The work done by me was:
	Sponsored by:   The FreeBSD Foundation

plus a few netbsd-specific tweaks from me.
fixes PR 50790.
2016-05-14 21:19:05 +00:00
christos
d7245a7577 include stdio.h for NULL 2016-04-28 11:45:02 +00:00
christos
f107a05fec remove incorrect comment 2016-04-28 11:38:41 +00:00
christos
4187cd60b2 Introduce dt_bootfile() and use it to avoid hard-coded kernel values. Fixes
dtrace hang for kernels not named netbsd.
2016-04-28 00:02:40 +00:00
christos
cb19fc18b8 Apply revision 274569 from FreeBSD:
Only compare visitation counters if they've both been set for the current
type graph walk.
2016-04-10 23:37:10 +00:00
christos
0aab751630 fix test. 2016-04-09 17:58:46 +00:00
christos
eabf50fc2e properly check for type id overflow 2016-04-09 17:51:50 +00:00
riastradh
989878b760 Add autounload logic to dtrace module prototype. 2016-04-09 15:18:48 +00:00
riastradh
51346d2825 Disable autoload if there are any profile probes. 2016-04-09 15:17:58 +00:00
riastradh
7256e74462 Prevent autounload while cyclics are still in use. 2016-04-09 14:50:08 +00:00
riastradh
0def3bc2ac Fix module name of dtrace_profile. 2016-04-04 05:15:07 +00:00
christos
ccd8517e35 write references as pointers XXX: 2016-03-18 17:11:04 +00:00
christos
8db25f0638 treat class line struct and union 2016-03-18 17:08:45 +00:00
christos
dd09ff20c8 Add support for c++ classes. 2016-03-18 17:07:23 +00:00
christos
63ffeeee3f spell restrict 2016-03-18 16:37:09 +00:00
christos
3e0f598a15 fix printf format 2016-03-18 16:32:03 +00:00
christos
714b78ffa0 Add references, comments. 2016-03-18 16:24:26 +00:00
christos
096b9466af add references. 2016-03-18 16:12:46 +00:00
christos
4a929da448 Add references. 2016-03-18 15:26:31 +00:00
christos
22c82d5b7b - Add processing for c++ references.
- Make sure we load the DIE that contains the types of array elements so
  we can resolve them later.
- Print t_id (die offsets) in hex.
2016-03-18 14:55:34 +00:00
christos
f15ed7b01a swap first, then write. 2016-03-18 13:23:00 +00:00
christos
cdd20e12b2 Avoid longjmp clobbering 2016-03-17 03:05:55 +00:00
joerg
fbba0696d4 Drop FreeBSD RCS ID, add NetBSD's. 2016-03-11 05:02:32 +00:00
christos
603f28d89d remove comment I used to test. 2016-03-01 21:45:10 +00:00
christos
f485d9df89 Build as PIC if requested 2016-03-01 21:30:40 +00:00
joerg
2f04f4ac92 Don't crash when relocbase was 0. 2016-03-01 21:09:17 +00:00
joerg
f2e47b4038 Merge r274564 from FreeBSD:
Fix a couple of bugs around the handling of structs and unions of size zero.
These would cause ctfconvert(1) to return an error when attempting to
resolve valid C types.
2016-03-01 14:52:14 +00:00
riastradh
a37e72fe0d Let MKCTF alone control whether we build libctf.
If MKDTRACE=yes MKCTF=no makes no sense, we can express that
somewhere else, not as a confusing set lists error at the end -- as
far as I know, there's no way to express MKDTRACE *or* MKCTF as a
condition for set list entries.
2016-02-29 17:46:54 +00:00
riastradh
9e84fa00b8 uint64_t can't be negative. 2016-02-29 16:19:20 +00:00
riastradh
a9ebcda81c Sprinkle some more zfs range lock reference count kasserts.
From a debugging session months ago, might as well share them.
2016-02-29 16:18:37 +00:00
riastradh
0ba9aea45e Cast int64_t to long long before printfing with %lld.
Not doing so broke something months ago and I forgot why.
2016-02-29 16:14:57 +00:00
riastradh
886308980e Add __printflike as appropriate. 2016-02-29 02:19:37 +00:00
christos
042ef76aff - Add missing ioctl.h
- Hide dprintf
2016-02-28 19:10:45 +00:00
christos
91780cc76d hook drti 2016-02-28 19:10:10 +00:00
joerg
f247f26769 Make the default CPP path a list and check if they are access(2)able.
If no match is found, fallback to looking for cpp in PATH.
2016-02-24 21:11:34 +00:00
joerg
5e22f35b4d Keep the if chain going. 2016-02-21 13:33:47 +00:00
joerg
486516affa Clang 3.7 and newer provide the array size via DW_AT_count, not via
DW_AT_upper_bound. Recognize the former as well as the latter.
2016-02-20 23:09:03 +00:00
joerg
4fa4785fa4 Don't nuke input on errors, it makes debugging a nightmare. 2016-02-20 22:08:44 +00:00
joerg
f9442aa68f Shuffle include order to avoid missing free on glibc. 2016-02-20 21:50:02 +00:00
christos
604231c4f0 - don't barf if the object does not have DWARF debug data.
- bump size of types to 1K to avoid string overflow
(both are needed for the new elftoolchain).
2016-02-20 02:39:47 +00:00
joerg
b0b911ff3e Don't use #pragma weak to implicitly change argument types. 2016-02-19 19:25:59 +00:00
joerg
e9ae0934d6 dmu_objset_rename is an alias for dsl_dataset_rename, but differs in
constness of the first argument. Newer clang checks for this and
prohibits it, so just be consistent by constifying the latter.
2016-02-19 19:25:22 +00:00
christos
c628030d37 reduce ifdefs 2016-02-13 21:37:12 +00:00
joerg
66313330f0 SDT is not a disease. Use const consistently. 2016-02-06 18:19:59 +00:00
riastradh
8e1ee72c35 Need nbtool_config.h for __dead in tools. 2016-02-05 20:39:50 +00:00
riastradh
32e3d5d381 Need nbtools_config.h for __unused.
Fixes

.../src/tools/ctfmerge/../../external/cddl/osnet/dist/tools/ctf/common/list.c:73:47: error: expected ';', ',' or ')' before '__unused'
 list_defcmp(void *d1, void *d2, void *private __unused)
                                               ^~~~~~~~

on non-NetBSD systems, where <assert.h>, <stdlib.h>, <stdio.h>, and
<sys/types.h> don't define __unused.
2016-02-05 19:56:00 +00:00
roy
871759c0d7 Fix the actual issue rather than silencing the warning.
Thanks to gson@
2016-02-05 10:14:49 +00:00
roy
f4a5ac8842 Add -Wno-format-nonliteral 2016-02-04 21:26:48 +00:00
roy
5baa0e8c89 dtrace needs Wno-format-security for clang it seems. 2016-02-04 20:06:29 +00:00