d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
639ec13157
freebsd-dev/sys/kern/imgact_elf.c

2765 lines
74 KiB
C
Raw Normal View History

First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*-
sys/kern: adoption of SPDX licensing ID tags. Mainly focus on files that use BSD 2-Clause license, however the tool I was using misidentified many licenses so this was mostly a manual - error prone - task. The Software Package Data Exchange (SPDX) group provides a specification to make it easier for automated tools to detect and summarize well known opensource licenses. We are gradually adopting the specification, noting that the tags are considered only advisory and do not, in any way, superceed or replace the license texts.
2017-11-27 15:20:12 +00:00
* SPDX-License-Identifier: BSD-3-Clause
*
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
* Copyright (c) 2017 Dell EMC
Correct copyright dates.
2018-07-30 07:01:00 +00:00
* Copyright (c) 2000-2001, 2003 David O'Brien
Convert files to UTF-8
2012-01-15 13:23:18 +00:00
* Copyright (c) 1995-1996 Søren Schmidt
Fix some rounding problems.. In some (fairly rare) situtaions it mapped one page too many, which caused obreak() to fail in vm_map_find() with ENOMEM because of the conflicting page.
1996-03-10 22:37:34 +00:00
* Copyright (c) 1996 Peter Wemm
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer
* in this position and unchanged.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
Fix typo in the BSD copyright: s/withough/without/ Spotted and suggested by: des MFC after: 3 weeks
2002-06-02 20:05:59 +00:00
* derived from this software without specific prior written permission
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
Use __FBSDID().
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Add some checks to ensure that Capsicum is behaving correctly, and add some more explicit comments about what's going on and what future maintainers need to do when e.g. adding a new operation to a sys_machdep.c. Approved by: mentor(rwatson), re(bz)
2011-06-30 10:56:02 +00:00
#include "opt_capsicum.h"
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/param.h>
Update kernel inclusions of capability.h to use capsicum.h instead; some further refinement is required as some device drivers intended to be portable over FreeBSD versions rely on __FreeBSD_version to decide whether to include capability.h. MFC after: 3 weeks
2014-03-16 10:55:57 +00:00
#include <sys/capsicum.h>
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
#include <sys/compressor.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/exec.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/fcntl.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/imgact.h>
#include <sys/imgact_elf.h>
Allow the kern.osrelease and kern.osreldate sysctl values to be set in a jail's creation parameters. This allows the kernel version to be reliably spoofed within the jail whether examined directly with sysctl or indirectly with the uname -r and -K options. The values can only be set at jail creation time, to eliminate the need for any locking when accessing the values via sysctl. The overridden values are inherited by nested jails (unless the config for the nested jails also overrides the values). There is no sanity or range checking, other than disallowing an empty release string or a zero release date, by design. The system administrator is trusted to set sane values. Setting values that are newer than the actual running kernel will likely cause compatibility problems. Differential Revision: https://reviews.freebsd.org/D1948 Relnotes: yes
2015-02-27 16:28:55 +00:00
#include <sys/jail.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/kernel.h>
Catch up to header include changes: - <sys/mutex.h> now requires <sys/systm.h> - <sys/mutex.h> and <sys/sx.h> now require <sys/lock.h>
2001-03-28 09:17:56 +00:00
#include <sys/lock.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/malloc.h>
Improve the MP safeness associated with the creation of symbolic links and the execution of ELF binaries. Two problems were found: 1) The link path wasn't tagged as being MP safe and thus was not properly protected. 2) The ELF interpreter vnode wasnt being locked in namei(9) and thus was insufficiently protected. This commit makes the following changes: -Sets the MPSAFE flag in NDINIT for symbolic link paths -Sets the MPSAFE flag in NDINIT and introduce a vfslocked variable which will be used to instruct VFS_UNLOCK_GIANT to unlock Giant if it has been picked up. -Drop in an assertion into vfs_lookup which ensures that if the MPSAFE flag is NOT set, that we have picked up giant. If not panic (if WITNESS compiled into the kernel). This should help us find conditions where vnode operations are in-sufficiently protected. This is a RELENG_6 candidate. Discussed with: jeff MFC after: 4 days
2005-09-15 15:03:48 +00:00
#include <sys/mount.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/mman.h>
Removed unnecessary #includes from <sys/imgact.h> so that it is self-sufficient and added explicit #includes where required.
1996-05-01 02:43:13 +00:00
#include <sys/namei.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/pioctl.h>
Removed unnecessary #includes from <sys/imgact.h> so that it is self-sufficient and added explicit #includes where required.
1996-05-01 02:43:13 +00:00
#include <sys/proc.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/procfs.h>
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
#include <sys/ptrace.h>
Add accounting for most of the memory-related resources. Sponsored by: The FreeBSD Foundation Reviewed by: kib (earlier version)
2011-04-05 20:23:59 +00:00
#include <sys/racct.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/resourcevar.h>
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
#include <sys/rwlock.h>
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
#include <sys/sbuf.h>
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
#include <sys/sf_buf.h>
Supply some useful information to the started image using ELF aux vectors. In particular, provide pagesize and pagesizes array, the canary value for SSP use, number of host CPUs and osreldate. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 08:55:45 +00:00
#include <sys/smp.h>
Move the include of <sys/systm.h> so that KTR gets a declaration for snprintf().
2000-09-10 13:54:52 +00:00
#include <sys/systm.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/signalvar.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/stat.h>
Convert the allproc and proctree locks from lockmgr locks to sx locks.
2001-03-28 11:52:56 +00:00
#include <sys/sx.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/syscall.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <sys/sysctl.h>
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
#include <sys/sysent.h>
Removed unnecessary #includes from <sys/imgact.h> so that it is self-sufficient and added explicit #includes where required.
1996-05-01 02:43:13 +00:00
#include <sys/vnode.h>
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
#include <sys/syslog.h>
#include <sys/eventhandler.h>
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
#include <sys/user.h>
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <vm/vm.h>
#include <vm/vm_kern.h>
#include <vm/vm_param.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
#include <vm/vm_object.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <vm/vm_extern.h>
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
#include <machine/elf.h>
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
#include <machine/md_var.h>
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
#define ELF_NOTE_ROUNDSIZE 4
Change our ELF binary branding to something more acceptable to the Binutils maintainers. After we established our branding method of writing upto 8 characters of the OS name into the ELF header in the padding; the Binutils maintainers and/or SCO (as USL) decided that instead the ELF header should grow two new fields -- EI_OSABI and EI_ABIVERSION. Each of these are an 8-bit unsigned integer. SCO has assigned official values for the EI_OSABI field. In addition to this, the Binutils maintainers and NetBSD decided that a better ELF branding method was to include ABI information in a ".note" ELF section. With this set of changes, we will now create ELF binaries branded using both "official" methods. Due to the complexity of adding a section to a binary, binaries branded with ``brandelf'' will only brand using the EI_OSABI method. Also due to the complexity of pulling a section out of an ELF file vs. poking around in the ELF header, our image activator only looks at the EI_OSABI header field. Note that a new kernel can still properly load old binaries except for Linux static binaries branded in our old method. * * For a short period of time, ``ld'' will also brand ELF binaries * using our old method. This is so people can still use kernel.old * with a new world. This support will be removed before 5.0-RELEASE, * and may not last anywhere upto the actual release. My expiration * time for this is about 6mo. *
2000-04-18 02:39:26 +00:00
#define OLD_EI_BRAND 8
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
static int __elfN(check_header)(const Elf_Ehdr *hdr);
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
static Elf_Brandinfo *__elfN(get_brandinfo)(struct image_params *imgp,
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
const char *interp, int32_t *osrel, uint32_t *fctl0);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
static int __elfN(load_file)(struct proc *p, const char *file, u_long *addr,
Remove sv_pagesize, originally introduced with r100384. In all of the architectures we have today, we always use PAGE_SIZE. While in theory one could define different things, none of the current architectures do, even the ones that have transitioned from 32-bit to 64-bit like i386 and arm. Some ancient mips binaries on other systems used 8k instead of 4k, but we don't support running those and likely never will due to their age and obscurity. Reviewed by: imp (who also contributed the commit message) Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19280
2019-03-01 16:16:38 +00:00
u_long *entry);
Consistently use vm_ooffset_t type for the vm object offset in elf_load_section. The values passed currently as vm_offset_t are phdr.p_offset, which have the native Elf word size. Since elf_load_section interprets them as the file offset, use vm object offset type. Noted and reviewed by: alc Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:36:43 +00:00
static int __elfN(load_section)(struct image_params *imgp, vm_ooffset_t offset,
Remove sv_pagesize, originally introduced with r100384. In all of the architectures we have today, we always use PAGE_SIZE. While in theory one could define different things, none of the current architectures do, even the ones that have transitioned from 32-bit to 64-bit like i386 and arm. Some ancient mips binaries on other systems used 8k instead of 4k, but we don't support running those and likely never will due to their age and obscurity. Reviewed by: imp (who also contributed the commit message) Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19280
2019-03-01 16:16:38 +00:00
caddr_t vmaddr, size_t memsz, size_t filsz, vm_prot_t prot);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
static int __CONCAT(exec_, __elfN(imgact))(struct image_params *imgp);
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
static bool __elfN(freebsd_trans_osrel)(const Elf_Note *note,
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
int32_t *osrel);
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
static bool kfreebsd_trans_osrel(const Elf_Note *note, int32_t *osrel);
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
static boolean_t __elfN(check_note)(struct image_params *imgp,
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
Elf_Brandnote *checknote, int32_t *osrel, uint32_t *fctl0);
Collect code to translate between vm_prot_t and p_flags into helper functions. MFC after: 1 week
2011-01-08 16:02:14 +00:00
static vm_prot_t __elfN(trans_prot)(Elf_Word);
static Elf_Word __elfN(untrans_prot)(vm_prot_t);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
SYSCTL_NODE(_kern, OID_AUTO, __CONCAT(elf, __ELF_WORD_SIZE), CTLFLAG_RW, 0,
"");
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
#define CORE_BUF_SIZE (16 * 1024)
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
- Provide backwards compatibility for kern.fallback_elf_brand. - Use the generic elf type macros in imgact_elf.h instead of ifdefing the entire contents of the header.
2003-01-05 03:48:14 +00:00
int __elfN(fallback_brand) = -1;
SYSCTL_INT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO,
Pull in r267961 and r267973 again. Fix for issues reported will follow.
2014-06-28 03:56:17 +00:00
fallback_brand, CTLFLAG_RWTUN, &__elfN(fallback_brand), 0,
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE)) " brand of last resort");
Fix multiple registration of the elf_legacy_coredump sysctl variable. The duplication is caused by the fact that imgact_elf.c is included by both imgact_elf32.c and imgact_elf64.c and both are compiled by default on ia64. Consequently, we have two seperate copies of the elf_legacy_coredump variable due to them being declared static, and two entries for the same sysctl in the linker set, both referencing the unique copy of the elf_legacy_coredump variable. Since the second sysctl cannot be registered, one of the elf_legacy_coredump variables can not be tuned (if ordering still holds, it's the ELF64 related one). The only solution is to create two different sysctl variables, just like the elf<32|64>_trace sysctl variables. This unfortunately is an (user) interface change, but unavoidable. Thus, on ELF32 platforms the sysctl variable is called elf32_legacy_coredump and on ELF64 platforms it is called elf64_legacy_coredump. Platforms that have both ELF formats have both sysctl variables. These variables should probably be retired sooner rather than later.
2002-12-21 01:15:39 +00:00
static int elf_legacy_coredump = 0;
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
SYSCTL_INT(_debug, OID_AUTO, __elfN(legacy_coredump), CTLFLAG_RW,
add description for debug.elf{32,64}_legacy_coredump sysctl Approved by: re (kib) MFC after: 1 week Sponsored by: The FreeBSD Foundation
2016-07-05 14:46:06 +00:00
&elf_legacy_coredump, 0,
"include all and only RW pages in core dumps");
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Finally, try to enable the nxstacks on amd64 and powerpc64 for both 64bit and 32bit ABIs. Also try to enable nxstacks for PAE/i386 when supported, and some variants of powerpc32. MFC after: 2 months (if ever)
2012-01-30 07:56:00 +00:00
int __elfN(nxstack) =
Add support for usermode (vdso-like) gettimeofday(2) and clock_gettime(2) on ARMv7 and ARMv8 systems which have architectural generic timer hardware. It is similar how the RDTSC timer is used in userspace on x86. Fix a permission problem where generic timer access from EL0 (or userspace on v7) was not properly initialized on APs. For ARMv7, mark the stack non-executable. The shared page is added for all arms (including ARMv8 64bit), and the signal trampoline code is moved to the page. Reviewed by: andrew Discussed with: emaste, mmel Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D4209
2015-12-07 12:20:26 +00:00
#if defined(__amd64__) || defined(__powerpc64__) /* both 64 and 32 bit */ || \
Enable non-executable stacks by default on RISC-V. Reviewed by: markj Sponsored by: DARPA Differential Revision: https://reviews.freebsd.org/D17878
2018-11-07 18:32:02 +00:00
(defined(__arm__) && __ARM_ARCH >= 7) || defined(__aarch64__) || \
defined(__riscv)
Finally, try to enable the nxstacks on amd64 and powerpc64 for both 64bit and 32bit ABIs. Also try to enable nxstacks for PAE/i386 when supported, and some variants of powerpc32. MFC after: 2 months (if ever)
2012-01-30 07:56:00 +00:00
1;
#else
0;
#endif
In elf image activator, read and apply the stack protection mode from PT_GNU_STACK program header, if present and enabled. Two new sysctls are provided, kern.elf32.nxstack and kern.elf64.nxstack, that allow to enable PT_GNU_STACK for ABIs of specified bitsize, if ABI decided to support shared page. Inform rtld about access mode of the stack initial mapping by AT_STACKPROT aux vector. At the moment, the default is disabled, waiting for the usermode support bits.
2011-01-08 16:30:59 +00:00
SYSCTL_INT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO,
nxstack, CTLFLAG_RW, &__elfN(nxstack), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE)) ": enable non-executable stack");
Port sysctl kern.elf32.read_exec from amd64 to i386. Make it more comprehensive on i386, by not setting nx bit for any mapping, not just adding PF_X to all kernel-loaded ELF segments. This is needed for the compatibility with older i386 programs that assume that read access implies exec, e.g. old X servers with hand-rolled module loader. Reported and tested by: bde Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-02-07 02:17:34 +00:00
#if __ELF_WORD_SIZE == 32 && (defined(__amd64__) || defined(__i386__))
Control the execution permission of the readable segments for i386 binaries on the amd64 and ia64 with the sysctl, instead of unconditionally enabling it. Reviewed by: marcel
2011-10-15 12:35:18 +00:00
int i386_read_exec = 0;
SYSCTL_INT(_kern_elf32, OID_AUTO, read_exec, CTLFLAG_RW, &i386_read_exec, 0,
"enable execution from readable segments");
#endif
Make non-ASLR pie base tunable. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-09-21 18:00:23 +00:00
static u_long __elfN(pie_base) = ET_DYN_LOAD_ADDR;
kern.elf{32,64}.pie_base sysctl: enforce page alignment. Requested by: rstone Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-09-21 20:03:17 +00:00
static int
sysctl_pie_base(SYSCTL_HANDLER_ARGS)
{
u_long val;
int error;
val = __elfN(pie_base);
error = sysctl_handle_long(oidp, &val, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if ((val & PAGE_MASK) != 0)
return (EINVAL);
__elfN(pie_base) = val;
return (0);
}
SYSCTL_PROC(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO, pie_base,
CTLTYPE_ULONG | CTLFLAG_MPSAFE | CTLFLAG_RW, NULL, 0,
sysctl_pie_base, "LU",
Make non-ASLR pie base tunable. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-09-21 18:00:23 +00:00
"PIE load base without randomization");
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
SYSCTL_NODE(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO, aslr, CTLFLAG_RW, 0,
"");
#define ASLR_NODE_OID __CONCAT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), _aslr)
static int __elfN(aslr_enabled) = 0;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, enable, CTLFLAG_RWTUN,
&__elfN(aslr_enabled), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
": enable address map randomization");
static int __elfN(pie_aslr_enabled) = 0;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, pie_enable, CTLFLAG_RWTUN,
&__elfN(pie_aslr_enabled), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
": enable address map randomization for PIE binaries");
static int __elfN(aslr_honor_sbrk) = 1;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, honor_sbrk, CTLFLAG_RW,
&__elfN(aslr_honor_sbrk), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE)) ": assume sbrk is used");
Make randomized stack gap between strings and pointers to argv/envs. This effectively makes the stack base on the csu _start entry randomized. The gap is enabled if ASLR is for the ABI is enabled, and then kern.elf{64,32}.aslr.stack_gap specify the max percentage of the initial stack size that can be wasted for gap. Setting it to zero disables the gap, and max is capped at 50%. Only amd64 for now. Reviewed by: cem, markj Discussed with: emaste MFC after: 2 weeks Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D21081
2019-07-31 20:23:10 +00:00
static int __elfN(aslr_stack_gap) = 3;
SYSCTL_INT(ASLR_NODE_OID, OID_AUTO, stack_gap, CTLFLAG_RW,
&__elfN(aslr_stack_gap), 0,
__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
": maximum percentage of main stack to waste on a random gap");
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
static Elf_Brandinfo *elf_brand_list[MAX_BRANDS];
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Remove trunc_page_ps() and round_page_ps() macros. This completes the undoing of r100384. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19680
2019-03-23 13:41:14 +00:00
#define aligned(a, t) (rounddown2((u_long)(a), sizeof(t)) == (u_long)(a))
Check for the program headers alignment of the ELF images before dereferencing. Unaligned access could cause panic on strict alignment architectures. Reviewed by: marcel, marius (also tested on sparc64, thanks !) MFC after: 3 days
2007-12-04 12:21:27 +00:00
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
static const char FREEBSD_ABI_VENDOR[] = "FreeBSD";
Elf_Brandnote __elfN(freebsd_brandnote) = {
.hdr.n_namesz = sizeof(FREEBSD_ABI_VENDOR),
.hdr.n_descsz = sizeof(int32_t),
Replace magic value ELF note type with NT_FREEBSD_ABI_TAG As of r291909 elf_common.h provides a definition. Suggested by: kib Sponsored by: The FreeBSD Foundation
2015-12-07 18:43:27 +00:00
.hdr.n_type = NT_FREEBSD_ABI_TAG,
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
.vendor = FREEBSD_ABI_VENDOR,
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
.flags = BN_TRANSLATE_OSREL,
.trans_osrel = __elfN(freebsd_trans_osrel)
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
};
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
static bool
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
__elfN(freebsd_trans_osrel)(const Elf_Note *note, int32_t *osrel)
{
uintptr_t p;
p = (uintptr_t)(note + 1);
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
p += roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE);
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
*osrel = *(const int32_t *)(p);
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
return (true);
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
}
static const char GNU_ABI_VENDOR[] = "GNU";
static int GNU_KFREEBSD_ABI_DESC = 3;
Elf_Brandnote __elfN(kfreebsd_brandnote) = {
.hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
.hdr.n_descsz = 16, /* XXX at least 16 */
.hdr.n_type = 1,
.vendor = GNU_ABI_VENDOR,
.flags = BN_TRANSLATE_OSREL,
.trans_osrel = kfreebsd_trans_osrel
};
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
static bool
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
kfreebsd_trans_osrel(const Elf_Note *note, int32_t *osrel)
{
const Elf32_Word *desc;
uintptr_t p;
p = (uintptr_t)(note + 1);
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
p += roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE);
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
desc = (const Elf32_Word *)p;
if (desc[0] != GNU_KFREEBSD_ABI_DESC)
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
return (false);
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
/*
* Debian GNU/kFreeBSD embed the earliest compatible kernel version
* (__FreeBSD_version: <major><two digit minor>Rxx) in the LSB way.
*/
*osrel = desc[1] * 100000 + desc[2] * 1000 + desc[3];
Use C99 boolean type for translate_osrel Migrate to modern types before creating MD Linuxolator bits for new architectures. Reviewed by: cem Sponsored by: Turing Robotic Industries Inc. Differential Revision: https://reviews.freebsd.org/D14676
2018-03-13 16:40:29 +00:00
return (true);
Fix handling of .note.ABI-tag section for GNU systems [1]. Handle GNU/Linux according to LSB Core Specification 4.0, Chapter 11. Object Format, 11.8. ABI note tag. Also check the first word of desc, not only name, according to glibc abi-tags specification to distinguish between Linux and kFreeBSD. Add explicit handling for Debian GNU/kFreeBSD, which runs on our kernels as well [2]. In {amd64,i386}/trap.c, when checking osrel of the current process, also check the ABI to not change the signal behaviour for Linux binary processes, now that we save an osrel version for all three from the lists above in struct proc [2]. These changes make it possible to run FreeBSD, Debian GNU/kFreeBSD and Linux binaries on the same machine again for at least i386 and amd64, and no longer break kFreeBSD which was detected as GNU(/Linux). PR: kern/135468 Submitted by: dchagin [1] (initial patch) Suggested by: kib [2] Tested by: Petr Salinger (Petr.Salinger seznam.cz) for kFreeBSD Reviewed by: kib MFC after: 3 days
2009-08-24 16:19:47 +00:00
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__elfN(insert_brand_entry)(Elf_Brandinfo *entry)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
int i;
Fixed white space around operators, casts and reserved words. Reviewed by: md5
2002-08-24 22:55:16 +00:00
for (i = 0; i < MAX_BRANDS; i++) {
Prepare kernel to take advantage of "branded" ELF binaries.
1996-10-16 17:51:08 +00:00
if (elf_brand_list[i] == NULL) {
elf_brand_list[i] = entry;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
break;
}
}
Print a warning in case we cannot add more brandinfo because we would overflow the MAX_BRANDS sized array. Reviewed by: kib MFC After: 1 month
2009-10-03 10:50:00 +00:00
if (i == MAX_BRANDS) {
printf("WARNING: %s: could not insert brandinfo entry: %p\n",
__func__, entry);
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (-1);
Print a warning in case we cannot add more brandinfo because we would overflow the MAX_BRANDS sized array. Reviewed by: kib MFC After: 1 month
2009-10-03 10:50:00 +00:00
}
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (0);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__elfN(remove_brand_entry)(Elf_Brandinfo *entry)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
int i;
Fixed white space around operators, casts and reserved words. Reviewed by: md5
2002-08-24 22:55:16 +00:00
for (i = 0; i < MAX_BRANDS; i++) {
Prepare kernel to take advantage of "branded" ELF binaries.
1996-10-16 17:51:08 +00:00
if (elf_brand_list[i] == entry) {
elf_brand_list[i] = NULL;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
break;
}
}
Prepare kernel to take advantage of "branded" ELF binaries.
1996-10-16 17:51:08 +00:00
if (i == MAX_BRANDS)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (-1);
return (0);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
Provide elf_brand_inuse() as a method an emulator can use to find out whether it is currently in use (which is kinda useful when it's about to unload itself: Lockups are never very much fun, are they?).
1999-02-04 12:42:39 +00:00
int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__elfN(brand_inuse)(Elf_Brandinfo *entry)
Provide elf_brand_inuse() as a method an emulator can use to find out whether it is currently in use (which is kinda useful when it's about to unload itself: Lockups are never very much fun, are they?).
1999-02-04 12:42:39 +00:00
{
struct proc *p;
Protect the following with a lockmgr lock: allproc zombproc pidhashtbl proc.p_list proc.p_hash nextpid Reviewed by: jhb Obtained from: BSD/OS and netbsd
2000-11-22 07:42:04 +00:00
int rval = FALSE;
Provide elf_brand_inuse() as a method an emulator can use to find out whether it is currently in use (which is kinda useful when it's about to unload itself: Lockups are never very much fun, are they?).
1999-02-04 12:42:39 +00:00
Convert the allproc and proctree locks from lockmgr locks to sx locks.
2001-03-28 11:52:56 +00:00
sx_slock(&allproc_lock);
Use FOREACH_PROC_IN_SYSTEM instead of using its unrolled form.
2007-01-17 14:58:53 +00:00
FOREACH_PROC_IN_SYSTEM(p) {
Protect the following with a lockmgr lock: allproc zombproc pidhashtbl proc.p_list proc.p_hash nextpid Reviewed by: jhb Obtained from: BSD/OS and netbsd
2000-11-22 07:42:04 +00:00
if (p->p_sysent == entry->sysvec) {
rval = TRUE;
break;
}
Provide elf_brand_inuse() as a method an emulator can use to find out whether it is currently in use (which is kinda useful when it's about to unload itself: Lockups are never very much fun, are they?).
1999-02-04 12:42:39 +00:00
}
Convert the allproc and proctree locks from lockmgr locks to sx locks.
2001-03-28 11:52:56 +00:00
sx_sunlock(&allproc_lock);
Provide elf_brand_inuse() as a method an emulator can use to find out whether it is currently in use (which is kinda useful when it's about to unload itself: Lockups are never very much fun, are they?).
1999-02-04 12:42:39 +00:00
Protect the following with a lockmgr lock: allproc zombproc pidhashtbl proc.p_list proc.p_hash nextpid Reviewed by: jhb Obtained from: BSD/OS and netbsd
2000-11-22 07:42:04 +00:00
return (rval);
Provide elf_brand_inuse() as a method an emulator can use to find out whether it is currently in use (which is kinda useful when it's about to unload itself: Lockups are never very much fun, are they?).
1999-02-04 12:42:39 +00:00
}
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
static Elf_Brandinfo *
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
__elfN(get_brandinfo)(struct image_params *imgp, const char *interp,
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
int32_t *osrel, uint32_t *fctl0)
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
{
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
const Elf_Ehdr *hdr = (const Elf_Ehdr *)imgp->image_header;
When matching brand to the ELF binary by notes, try to find a brand with interpreter name exactly matching one wanted by the binary. If no such brand exists, return first brand which accepted the binary by note. The change fixes a regression after r292749, where e.g. our two ia32 compat brands, ia32_brand_info and ia32_brand_oinfo, only differ by the interpeter path and binary matches to a brand by linkage order. Then old binaries which require /usr/libexec/ld-elf.so.1 but matched against ia32_brand_info with interp_path /libexec/ld-elf.so.1, were considered requiring non-standard interpreter name, and magic to force ld-elf32.so.1 did not happen. Note that it might make sense to apply the same selection of brands for other matching criteria, SCO EI_OSABI and 3.x string. Reported and tested by: dwmalone Sponsored by: The FreeBSD Foundation MFC after: 3 days
2016-02-04 20:55:49 +00:00
Elf_Brandinfo *bi, *bi_m;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
boolean_t ret;
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
int i, interp_name_len;
Fix branding after r345661. In particular, elf32 FreeBSD binaries were not executed on LP64 hosts. The interp_name_len value should account for the nul terminator. This is needed for strncmp()s in brand checking code to work. Reported by: andreast Sponsored by: The FreeBSD Foundation MFC after: 12 days (together with r345661)
2019-03-30 16:58:51 +00:00
interp_name_len = interp != NULL ? strlen(interp) + 1 : 0;
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
/*
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
* We support four types of branding -- (1) the ELF EI_OSABI field
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
* that SCO added to the ELF spec, (2) FreeBSD 3.x's traditional string
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
* branding w/in the ELF header, (3) path of the `interp_path'
* field, and (4) the ".note.ABI-tag" ELF section.
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
*/
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
/* Look for an ".note.ABI-tag" ELF section */
When matching brand to the ELF binary by notes, try to find a brand with interpreter name exactly matching one wanted by the binary. If no such brand exists, return first brand which accepted the binary by note. The change fixes a regression after r292749, where e.g. our two ia32 compat brands, ia32_brand_info and ia32_brand_oinfo, only differ by the interpeter path and binary matches to a brand by linkage order. Then old binaries which require /usr/libexec/ld-elf.so.1 but matched against ia32_brand_info with interp_path /libexec/ld-elf.so.1, were considered requiring non-standard interpreter name, and magic to force ld-elf32.so.1 did not happen. Note that it might make sense to apply the same selection of brands for other matching criteria, SCO EI_OSABI and 3.x string. Reported and tested by: dwmalone Sponsored by: The FreeBSD Foundation MFC after: 3 days
2016-02-04 20:55:49 +00:00
bi_m = NULL;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
Make sure FreeBSD binaries without .note.ABI-tag section work correctly and do not match a colliding Debian GNU/kFreeBSD brandinfo statements. For this mark the Debian GNU/kFreeBSD brandinfo that it must have an .note.ABI-tag section and ignore the old EI_OSABI brandinfo when comparing a possibly colliding set of options. Due to SYSINIT we add the brandinfo in a non-deterministic order, so native FreeBSD is not always first. We may want to consider to force native FreeBSD to come first as well. The only way a problem could currently be noticed is when running an i386 binary without the .note.ABI-tag on amd64 and the Debian GNU/kFreeBSD brandinfo was matched first, as the fallback to ld-elf32.so.1 does not exist in that case. Reported and tested by: ticso In collaboration with: kib MFC after: 3 days
2009-08-30 14:38:17 +00:00
if (bi == NULL)
continue;
Update r315753 with the proper flag name. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:28:13 +00:00
if (interp != NULL && (bi->flags & BI_BRAND_ONLY_STATIC) != 0)
Add a flag BI_BRAND_ONLY_STATIC to specify that the brand only matches static binaries. Interpretation of the 'static' there is that the binary must not specify an interpreter. In particular, shared objects are matched by the brand if BI_CAN_EXEC_DYN is also set. This improves precision of the brand matching, which should eliminate surprises due to brand ordering. Revert r315701. Discussed with and tested by: ed (previous version) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:23:01 +00:00
continue;
Make sure FreeBSD binaries without .note.ABI-tag section work correctly and do not match a colliding Debian GNU/kFreeBSD brandinfo statements. For this mark the Debian GNU/kFreeBSD brandinfo that it must have an .note.ABI-tag section and ignore the old EI_OSABI brandinfo when comparing a possibly colliding set of options. Due to SYSINIT we add the brandinfo in a non-deterministic order, so native FreeBSD is not always first. We may want to consider to force native FreeBSD to come first as well. The only way a problem could currently be noticed is when running an i386 binary without the .note.ABI-tag on amd64 and the Debian GNU/kFreeBSD brandinfo was matched first, as the fallback to ld-elf32.so.1 does not exist in that case. Reported and tested by: ticso In collaboration with: kib MFC after: 3 days
2009-08-30 14:38:17 +00:00
if (hdr->e_machine == bi->machine && (bi->flags &
(BI_BRAND_NOTE|BI_BRAND_NOTE_MANDATORY)) != 0) {
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
ret = __elfN(check_note)(imgp, bi->brand_note, osrel,
fctl0);
Missed header_supported call from r291020: make really, really sure the brand likes the executable.
2015-12-01 17:00:31 +00:00
/* Give brand a chance to veto check_note's guess */
if (ret && bi->header_supported)
ret = bi->header_supported(imgp);
When matching brand to the ELF binary by notes, try to find a brand with interpreter name exactly matching one wanted by the binary. If no such brand exists, return first brand which accepted the binary by note. The change fixes a regression after r292749, where e.g. our two ia32 compat brands, ia32_brand_info and ia32_brand_oinfo, only differ by the interpeter path and binary matches to a brand by linkage order. Then old binaries which require /usr/libexec/ld-elf.so.1 but matched against ia32_brand_info with interp_path /libexec/ld-elf.so.1, were considered requiring non-standard interpreter name, and magic to force ld-elf32.so.1 did not happen. Note that it might make sense to apply the same selection of brands for other matching criteria, SCO EI_OSABI and 3.x string. Reported and tested by: dwmalone Sponsored by: The FreeBSD Foundation MFC after: 3 days
2016-02-04 20:55:49 +00:00
/*
* If note checker claimed the binary, but the
* interpreter path in the image does not
* match default one for the brand, try to
* search for other brands with the same
* interpreter. Either there is better brand
* with the right interpreter, or, failing
* this, we return first brand which accepted
* our note and, optionally, header.
*/
A followup to r315749, two more places where brand->interp_path was accessed unconditionally. Reported by: se Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-30 04:21:02 +00:00
if (ret && bi_m == NULL && interp != NULL &&
(bi->interp_path == NULL ||
(strlen(bi->interp_path) + 1 != interp_name_len ||
strncmp(interp, bi->interp_path, interp_name_len)
!= 0))) {
When matching brand to the ELF binary by notes, try to find a brand with interpreter name exactly matching one wanted by the binary. If no such brand exists, return first brand which accepted the binary by note. The change fixes a regression after r292749, where e.g. our two ia32 compat brands, ia32_brand_info and ia32_brand_oinfo, only differ by the interpeter path and binary matches to a brand by linkage order. Then old binaries which require /usr/libexec/ld-elf.so.1 but matched against ia32_brand_info with interp_path /libexec/ld-elf.so.1, were considered requiring non-standard interpreter name, and magic to force ld-elf32.so.1 did not happen. Note that it might make sense to apply the same selection of brands for other matching criteria, SCO EI_OSABI and 3.x string. Reported and tested by: dwmalone Sponsored by: The FreeBSD Foundation MFC after: 3 days
2016-02-04 20:55:49 +00:00
bi_m = bi;
ret = 0;
}
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
if (ret)
return (bi);
}
}
When matching brand to the ELF binary by notes, try to find a brand with interpreter name exactly matching one wanted by the binary. If no such brand exists, return first brand which accepted the binary by note. The change fixes a regression after r292749, where e.g. our two ia32 compat brands, ia32_brand_info and ia32_brand_oinfo, only differ by the interpeter path and binary matches to a brand by linkage order. Then old binaries which require /usr/libexec/ld-elf.so.1 but matched against ia32_brand_info with interp_path /libexec/ld-elf.so.1, were considered requiring non-standard interpreter name, and magic to force ld-elf32.so.1 did not happen. Note that it might make sense to apply the same selection of brands for other matching criteria, SCO EI_OSABI and 3.x string. Reported and tested by: dwmalone Sponsored by: The FreeBSD Foundation MFC after: 3 days
2016-02-04 20:55:49 +00:00
if (bi_m != NULL)
return (bi_m);
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
/* If the executable has a brand, search for it in the brand list. */
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
Add a flag BI_BRAND_ONLY_STATIC to specify that the brand only matches static binaries. Interpretation of the 'static' there is that the binary must not specify an interpreter. In particular, shared objects are matched by the brand if BI_CAN_EXEC_DYN is also set. This improves precision of the brand matching, which should eliminate surprises due to brand ordering. Revert r315701. Discussed with and tested by: ed (previous version) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:23:01 +00:00
if (bi == NULL || (bi->flags & BI_BRAND_NOTE_MANDATORY) != 0 ||
Update r315753 with the proper flag name. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:28:13 +00:00
(interp != NULL && (bi->flags & BI_BRAND_ONLY_STATIC) != 0))
Make sure FreeBSD binaries without .note.ABI-tag section work correctly and do not match a colliding Debian GNU/kFreeBSD brandinfo statements. For this mark the Debian GNU/kFreeBSD brandinfo that it must have an .note.ABI-tag section and ignore the old EI_OSABI brandinfo when comparing a possibly colliding set of options. Due to SYSINIT we add the brandinfo in a non-deterministic order, so native FreeBSD is not always first. We may want to consider to force native FreeBSD to come first as well. The only way a problem could currently be noticed is when running an i386 binary without the .note.ABI-tag on amd64 and the Debian GNU/kFreeBSD brandinfo was matched first, as the fallback to ld-elf32.so.1 does not exist in that case. Reported and tested by: ticso In collaboration with: kib MFC after: 3 days
2009-08-30 14:38:17 +00:00
continue;
if (hdr->e_machine == bi->machine &&
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
(hdr->e_ident[EI_OSABI] == bi->brand ||
Don't require the presence of the compat_3_brand. The existing ELF image activator requires the brandinfo to provide such a string unconditionally, even if the executable format in question doesn't use this type of branding. Skip matching when it's a null pointer. Reviewed by: kib MFC after: 2 weeks
2017-03-23 14:09:45 +00:00
(bi->compat_3_brand != NULL &&
Require whole brand string matching for old Elf branding. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:37:35 +00:00
strcmp((const char *)&hdr->e_ident[OLD_EI_BRAND],
Don't require the presence of the compat_3_brand. The existing ELF image activator requires the brandinfo to provide such a string unconditionally, even if the executable format in question doesn't use this type of branding. Skip matching when it's a null pointer. Reviewed by: kib MFC after: 2 weeks
2017-03-23 14:09:45 +00:00
bi->compat_3_brand) == 0))) {
Extend r270123 to run the brand info's header_supported() routine for branded as well as unbranded binaries. This will be required to add support for the new ELFv2 ABI on powerpc64, which is distinguished from ELFv1 by the contents of the ELF header's flags field. Reviewed by: imp MFC after: 2 weeks
2015-11-18 17:03:22 +00:00
/* Looks good, but give brand a chance to veto */
Always give ELF brands a chance to veto a match. If a brand provides a header_supported hook, check it when trying to find a brand based on a matching interpreter as well as in the final loop for the fallback brand. Previously a brand might reject a binary via a header_supported hook in one of the earlier loops, but still be chosen by one of these later loops. Reviewed by: kib Obtained from: CheriBSD MFC after: 2 weeks Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D13945
2018-02-05 23:27:42 +00:00
if (bi->header_supported == NULL ||
When selecting brand based on old Elf branding, prefer the brand which interpreter exactly matches the one requested by the activated image. This change applies r295277, which did the same for note branding, to the old brand selection, with the same reasoning of fixing compat32 interpreter substitution. PR: 211837 Reported by: kenji@kens.fm Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:38:25 +00:00
bi->header_supported(imgp)) {
/*
* Again, prefer strictly matching
* interpreter path.
*/
Adjust r314851 to not require every brand to specify interpreter path. Reported and tested by: ed Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:06:48 +00:00
if (interp_name_len == 0 &&
bi->interp_path == NULL)
return (bi);
if (bi->interp_path != NULL &&
strlen(bi->interp_path) + 1 ==
When selecting brand based on old Elf branding, prefer the brand which interpreter exactly matches the one requested by the activated image. This change applies r295277, which did the same for note branding, to the old brand selection, with the same reasoning of fixing compat32 interpreter substitution. PR: 211837 Reported by: kenji@kens.fm Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:38:25 +00:00
interp_name_len && strncmp(interp,
bi->interp_path, interp_name_len) == 0)
return (bi);
if (bi_m == NULL)
bi_m = bi;
}
Extend r270123 to run the brand info's header_supported() routine for branded as well as unbranded binaries. This will be required to add support for the new ELFv2 ABI on powerpc64, which is distinguished from ELFv1 by the contents of the ELF header's flags field. Reviewed by: imp MFC after: 2 weeks
2015-11-18 17:03:22 +00:00
}
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
}
When selecting brand based on old Elf branding, prefer the brand which interpreter exactly matches the one requested by the activated image. This change applies r295277, which did the same for note branding, to the old brand selection, with the same reasoning of fixing compat32 interpreter substitution. PR: 211837 Reported by: kenji@kens.fm Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:38:25 +00:00
if (bi_m != NULL)
return (bi_m);
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
Expand the elf brandelf infrastructure to give access to the whole ELF header (Elf_Ehdr) to determine if a particular interpretor wants to accept it or not. Use this mechanism to filter EABI arm on OABI arm kernels, and vice versa. This method could also be used to implement OABI on EABI arm kernels, if desired, or to allow a single mips kernel to run o32, n32 and n64 binaries. Differential Revision: https://reviews.freebsd.org/D609
2014-08-18 02:44:56 +00:00
/* No known brand, see if the header is recognized by any brand */
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
if (bi == NULL || bi->flags & BI_BRAND_NOTE_MANDATORY ||
bi->header_supported == NULL)
continue;
if (hdr->e_machine == bi->machine) {
ret = bi->header_supported(imgp);
if (ret)
return (bi);
}
}
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
/* Lacking a known brand, search for a recognized interpreter. */
if (interp != NULL) {
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
Update r315753 with the proper flag name. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:28:13 +00:00
if (bi == NULL || (bi->flags &
(BI_BRAND_NOTE_MANDATORY | BI_BRAND_ONLY_STATIC))
!= 0)
Make sure FreeBSD binaries without .note.ABI-tag section work correctly and do not match a colliding Debian GNU/kFreeBSD brandinfo statements. For this mark the Debian GNU/kFreeBSD brandinfo that it must have an .note.ABI-tag section and ignore the old EI_OSABI brandinfo when comparing a possibly colliding set of options. Due to SYSINIT we add the brandinfo in a non-deterministic order, so native FreeBSD is not always first. We may want to consider to force native FreeBSD to come first as well. The only way a problem could currently be noticed is when running an i386 binary without the .note.ABI-tag on amd64 and the Debian GNU/kFreeBSD brandinfo was matched first, as the fallback to ld-elf32.so.1 does not exist in that case. Reported and tested by: ticso In collaboration with: kib MFC after: 3 days
2009-08-30 14:38:17 +00:00
continue;
if (hdr->e_machine == bi->machine &&
A followup to r315749, two more places where brand->interp_path was accessed unconditionally. Reported by: se Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-30 04:21:02 +00:00
bi->interp_path != NULL &&
Fix several reads beyond the mapped first page of the binary in the ELF parser. Specifically, do not allow note reader and interpreter path comparision in the brandelf code to read past end of the page. This may happen if specially crafter ELF image is activated. Submitted by: Lukasz Wojcik <lukasz.wojcik zoho com> MFC after: 3 days
2012-07-19 11:15:53 +00:00
/* ELF image p_filesz includes terminating zero */
strlen(bi->interp_path) + 1 == interp_name_len &&
strncmp(interp, bi->interp_path, interp_name_len)
Always give ELF brands a chance to veto a match. If a brand provides a header_supported hook, check it when trying to find a brand based on a matching interpreter as well as in the final loop for the fallback brand. Previously a brand might reject a binary via a header_supported hook in one of the earlier loops, but still be chosen by one of these later loops. Reviewed by: kib Obtained from: CheriBSD MFC after: 2 weeks Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D13945
2018-02-05 23:27:42 +00:00
== 0 && (bi->header_supported == NULL ||
bi->header_supported(imgp)))
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
return (bi);
}
}
/* Lacking a recognized interpreter, try the default brand */
for (i = 0; i < MAX_BRANDS; i++) {
bi = elf_brand_list[i];
Add a flag BI_BRAND_ONLY_STATIC to specify that the brand only matches static binaries. Interpretation of the 'static' there is that the binary must not specify an interpreter. In particular, shared objects are matched by the brand if BI_CAN_EXEC_DYN is also set. This improves precision of the brand matching, which should eliminate surprises due to brand ordering. Revert r315701. Discussed with and tested by: ed (previous version) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:23:01 +00:00
if (bi == NULL || (bi->flags & BI_BRAND_NOTE_MANDATORY) != 0 ||
Update r315753 with the proper flag name. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-22 22:28:13 +00:00
(interp != NULL && (bi->flags & BI_BRAND_ONLY_STATIC) != 0))
Make sure FreeBSD binaries without .note.ABI-tag section work correctly and do not match a colliding Debian GNU/kFreeBSD brandinfo statements. For this mark the Debian GNU/kFreeBSD brandinfo that it must have an .note.ABI-tag section and ignore the old EI_OSABI brandinfo when comparing a possibly colliding set of options. Due to SYSINIT we add the brandinfo in a non-deterministic order, so native FreeBSD is not always first. We may want to consider to force native FreeBSD to come first as well. The only way a problem could currently be noticed is when running an i386 binary without the .note.ABI-tag on amd64 and the Debian GNU/kFreeBSD brandinfo was matched first, as the fallback to ld-elf32.so.1 does not exist in that case. Reported and tested by: ticso In collaboration with: kib MFC after: 3 days
2009-08-30 14:38:17 +00:00
continue;
if (hdr->e_machine == bi->machine &&
Always give ELF brands a chance to veto a match. If a brand provides a header_supported hook, check it when trying to find a brand based on a matching interpreter as well as in the final loop for the fallback brand. Previously a brand might reject a binary via a header_supported hook in one of the earlier loops, but still be chosen by one of these later loops. Reviewed by: kib Obtained from: CheriBSD MFC after: 2 weeks Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D13945
2018-02-05 23:27:42 +00:00
__elfN(fallback_brand) == bi->brand &&
(bi->header_supported == NULL ||
bi->header_supported(imgp)))
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
return (bi);
}
return (NULL);
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
static int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__elfN(check_header)(const Elf_Ehdr *hdr)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
Do not blow up when we walk off the end of the brands list. Found by: kris, jake
2002-09-08 02:17:44 +00:00
Elf_Brandinfo *bi;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
int i;
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
if (!IS_ELF(*hdr) ||
hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS ||
hdr->e_ident[EI_DATA] != ELF_TARG_DATA ||
Verify more bits of the ELF header: the program header table entry size and the ELF version. Also, avoid a potential integer overflow when determining whether the ELF header fits entirely within the first page. Reviewed by: jdp A panic when attempting to execute an ELF binary with a bogus program header table entry size was Reported by: Christer Öberg <christer.oberg@texonet.com>
2004-03-18 16:33:05 +00:00
hdr->e_ident[EI_VERSION] != EV_CURRENT ||
hdr->e_phentsize != sizeof(Elf_Phdr) ||
hdr->e_version != ELF_TARG_VER)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (ENOEXEC);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
/*
* Make sure we have at least one brand for this machine.
*/
Fixed white space around operators, casts and reserved words. Reviewed by: md5
2002-08-24 22:55:16 +00:00
for (i = 0; i < MAX_BRANDS; i++) {
Do not blow up when we walk off the end of the brands list. Found by: kris, jake
2002-09-08 02:17:44 +00:00
bi = elf_brand_list[i];
if (bi != NULL && bi->machine == hdr->e_machine)
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
break;
}
if (i == MAX_BRANDS)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (ENOEXEC);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (0);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
static int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__elfN(map_partial)(vm_map_t map, vm_object_t object, vm_ooffset_t offset,
Eliminate an unneeded (vm_prot_t) parameter from two functions. Eliminate unnecessary uses of a local variable. Reviewed by: tegge
2005-12-20 23:42:18 +00:00
vm_offset_t start, vm_offset_t end, vm_prot_t prot)
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
{
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
struct sf_buf *sf;
int error;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
vm_offset_t off;
/*
* Create the page if it doesn't exist yet. Ignore errors.
*/
Instead of direct use of vm_map_insert(), call vm_map_fixed(MAP_CHECK_EXCL). This KPI explicitely indicates the intent of creating the mapping at the fixed address, and incorporates the map locking into the callee. Suggested and reviewed by: alc Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-06 14:09:54 +00:00
vm_map_fixed(map, NULL, 0, trunc_page(start), round_page(end) -
trunc_page(start), VM_PROT_ALL, VM_PROT_ALL, MAP_CHECK_EXCL);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
/*
* Find the page from the underlying object.
*/
Style and punctuation fixes. Reviewed by: kib MFC after: 3 days
2017-03-05 23:59:04 +00:00
if (object != NULL) {
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
sf = vm_imgact_map_page(object, offset);
if (sf == NULL)
return (KERN_FAILURE);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
off = offset - trunc_page(offset);
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
error = copyout((caddr_t)sf_buf_kva(sf) + off, (caddr_t)start,
Fixed placement of operators. Wrapped long lines.
2002-08-25 20:48:45 +00:00
end - start);
Revert r253939: We cannot busy a page before doing pagefaults. Infact, it can deadlock against vnode lock, as it tries to vget(). Other functions, right now, have an opposite lock ordering, like vm_object_sync(), which acquires the vnode lock first and then sleeps on the busy mechanism. Before this patch is reinserted we need to break this ordering. Sponsored by: EMC / Isilon storage division Reported by: kib
2013-08-05 08:55:35 +00:00
vm_imgact_unmap_page(sf);
Style. Reviewed by: alc Sponsored by: The FreeBSD Foundation MFC after: 3 days
2017-03-02 17:35:13 +00:00
if (error != 0)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (KERN_FAILURE);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (KERN_SUCCESS);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
static int
When deallocating the vm object in elf_map_insert() due to vm_map_insert() failure, drop the vnode lock around the call to vm_object_deallocate(). Since the deallocated object is the vm object of the vnode, we might get the vnode lock recursion there. In fact, it is almost impossible to make vm_map_insert() failing there on stock kernel. Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-01 10:22:07 +00:00
__elfN(map_insert)(struct image_params *imgp, vm_map_t map, vm_object_t object,
vm_ooffset_t offset, vm_offset_t start, vm_offset_t end, vm_prot_t prot,
int cow)
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
{
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
struct sf_buf *sf;
vm_offset_t off;
Fix ia32 compat on ia64. Recent ia64 MD changes caused the garbage on the stack to be changed in a way incompatible with elf32_map_insert() where we used data_buf without initializing it for when the partial mapping resulting in a misaligned image (typical when the page size implied by the image is not the same as the page size in use by the kernel). Since data_buf is passed by reference to vm_map_find(), the compiler cannot warn about it. While here, move all local variables to the top of the function.
2003-05-31 19:55:05 +00:00
vm_size_t sz;
When deallocating the vm object in elf_map_insert() due to vm_map_insert() failure, drop the vnode lock around the call to vm_object_deallocate(). Since the deallocated object is the vm object of the vnode, we might get the vnode lock recursion there. In fact, it is almost impossible to make vm_map_insert() failing there on stock kernel. Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-01 10:22:07 +00:00
int error, locked, rv;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
if (start != trunc_page(start)) {
Fixed most indentation bugs.
2002-08-25 22:36:52 +00:00
rv = __elfN(map_partial)(map, object, offset, start,
Eliminate an unneeded (vm_prot_t) parameter from two functions. Eliminate unnecessary uses of a local variable. Reviewed by: tegge
2005-12-20 23:42:18 +00:00
round_page(start), prot);
Style and punctuation fixes. Reviewed by: kib MFC after: 3 days
2017-03-05 23:59:04 +00:00
if (rv != KERN_SUCCESS)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (rv);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
offset += round_page(start) - start;
start = round_page(start);
}
if (end != round_page(end)) {
Fixed most indentation bugs.
2002-08-25 22:36:52 +00:00
rv = __elfN(map_partial)(map, object, offset +
Eliminate an unneeded (vm_prot_t) parameter from two functions. Eliminate unnecessary uses of a local variable. Reviewed by: tegge
2005-12-20 23:42:18 +00:00
trunc_page(end) - start, trunc_page(end), end, prot);
Style and punctuation fixes. Reviewed by: kib MFC after: 3 days
2017-03-05 23:59:04 +00:00
if (rv != KERN_SUCCESS)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (rv);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
end = trunc_page(end);
}
Simplify the control flow and tidy up a comment in map_insert. In collaboration with: kib MFC after: 1 week
2017-03-11 18:57:13 +00:00
if (start >= end)
return (KERN_SUCCESS);
if ((offset & PAGE_MASK) != 0) {
/*
* The mapping is not page aligned. This means that we have
* to copy the data.
*/
rv = vm_map_fixed(map, NULL, 0, start, end - start,
prot | VM_PROT_WRITE, VM_PROT_ALL, MAP_CHECK_EXCL);
if (rv != KERN_SUCCESS)
return (rv);
if (object == NULL)
return (KERN_SUCCESS);
for (; start < end; start += sz) {
sf = vm_imgact_map_page(object, offset);
if (sf == NULL)
return (KERN_FAILURE);
off = offset - trunc_page(offset);
sz = end - start;
if (sz > PAGE_SIZE - off)
sz = PAGE_SIZE - off;
error = copyout((caddr_t)sf_buf_kva(sf) + off,
(caddr_t)start, sz);
vm_imgact_unmap_page(sf);
if (error != 0)
return (KERN_FAILURE);
offset += sz;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
} else {
Simplify the control flow and tidy up a comment in map_insert. In collaboration with: kib MFC after: 1 week
2017-03-11 18:57:13 +00:00
vm_object_reference(object);
rv = vm_map_fixed(map, object, offset, start, end - start,
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
prot, VM_PROT_ALL, cow | MAP_CHECK_EXCL |
(object != NULL ? MAP_VN_EXEC : 0));
Simplify the control flow and tidy up a comment in map_insert. In collaboration with: kib MFC after: 1 week
2017-03-11 18:57:13 +00:00
if (rv != KERN_SUCCESS) {
locked = VOP_ISLOCKED(imgp->vp);
VOP_UNLOCK(imgp->vp, 0);
vm_object_deallocate(object);
vn_lock(imgp->vp, locked | LK_RETRY);
return (rv);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
} else if (object != NULL) {
MPASS(imgp->vp->v_object == object);
VOP_SET_TEXT_CHECKED(imgp->vp);
Simplify the control flow and tidy up a comment in map_insert. In collaboration with: kib MFC after: 1 week
2017-03-11 18:57:13 +00:00
}
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
Simplify the control flow and tidy up a comment in map_insert. In collaboration with: kib MFC after: 1 week
2017-03-11 18:57:13 +00:00
return (KERN_SUCCESS);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
static int
Consistently use vm_ooffset_t type for the vm object offset in elf_load_section. The values passed currently as vm_offset_t are phdr.p_offset, which have the native Elf word size. Since elf_load_section interprets them as the file offset, use vm object offset type. Noted and reviewed by: alc Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:36:43 +00:00
__elfN(load_section)(struct image_params *imgp, vm_ooffset_t offset,
Remove sv_pagesize, originally introduced with r100384. In all of the architectures we have today, we always use PAGE_SIZE. While in theory one could define different things, none of the current architectures do, even the ones that have transitioned from 32-bit to 64-bit like i386 and arm. Some ancient mips binaries on other systems used 8k instead of 4k, but we don't support running those and likely never will due to their age and obscurity. Reviewed by: imp (who also contributed the commit message) Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19280
2019-03-01 16:16:38 +00:00
caddr_t vmaddr, size_t memsz, size_t filsz, vm_prot_t prot)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
struct sf_buf *sf;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
size_t map_len;
Improve abstraction. Eliminate direct access by elf*_load_section() to an OBJT_VNODE-specific field of the vm object. The same information can be just as easily obtained from the struct vattr that is in struct image_params if the latter is passed to elf*_load_section(). Moreover, by replacing the vmspace and vm object parameters to elf*_load_section() with a struct image_params parameter, we actually reduce the size of the object code. In collaboration with: kib
2012-01-17 00:27:32 +00:00
vm_map_t map;
vm_object_t object;
Remove a redundant offset computation in elf_load_section(). With r344705 the offset is always zero. Submitted by: Wuyang Chung <wuyang.chung1@gmail.com>
2019-07-24 15:18:05 +00:00
vm_offset_t map_addr;
Change the way ELF coredumps are handled. Instead of unconditionally skipping read-only pages, which can result in valuable non-text-related data not getting dumped, the ELF loader and the dynamic loader now mark read-only text pages NOCORE and the coredump code only checks (primarily) for complete inaccessibility of the page or NOCORE being set. Certain applications which map large amounts of read-only data will produce much larger cores. A new sysctl has been added, debug.elf_legacy_coredump, which will revert to the old behavior. This commit represents collaborative work by all parties involved. The PR contains a program demonstrating the problem. PR: kern/45994 Submitted by: "Peter Edwards" <pmedwards@eircom.net>, Archie Cobbs <archie@dellroad.org> Reviewed by: jdp, dillon MFC after: 7 days
2002-12-16 19:24:43 +00:00
int error, rv, cow;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
size_t copy_len;
Consistently use vm_ooffset_t type for the vm object offset in elf_load_section. The values passed currently as vm_offset_t are phdr.p_offset, which have the native Elf word size. Since elf_load_section interprets them as the file offset, use vm object offset type. Noted and reviewed by: alc Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-07 13:36:43 +00:00
vm_ooffset_t file_addr;
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
Solve the problem where it is possible to get the kernel stuck in a loop down in pmap_init_pt(). A subtraction causes the number of pages to become negative, that was assigned to an unsigned variable, and there is a lot of iteration. The bug is due to the ELF image activator not properly checking for its files being the correct size as specified by the ELF header. The solution is to check that the header doesn't ask for part of a file when that part of the file doesn't exist. Make sure to set VEXEC at the proper times to make the executables immutable (remove race conditions). Also, the ELF format specifiies header entries that allow embedding of other executables (hence how ld-elf.so.1 gets loaded, but not the same as loading shared libraries), so those executables need to be set VEXEC, too, so they're immutable. Reviewed by: peter
2000-07-23 06:49:46 +00:00
/*
* It's necessary to fail if the filsz + offset taken from the
* header is greater than the actual file pager object's size.
* If we were to allow this, then the vm_map_find() below would
* walk right off the end of the file object and into the ether.
*
* While I'm here, might as well check for something else that
* is invalid: filsz cannot be greater than memsz.
*/
Accept linkers representation for ELF segments with zero on-disk length. For such segments, GNU bfd linker writes knowingly incorrect value into the the file offset field of the program header entry, with the motivation that file should not be mapped for creation of this segment at all. Relax checks for the ELF structure validity when on-disk segment length is zero, and explicitely set mapping length to zero for such segments to avoid validating rounding arithmetic. PR: 217610 Reported by: Robert Clausecker <fuz@fuz.su> Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-12 13:51:13 +00:00
if ((filsz != 0 && (off_t)filsz + offset > imgp->attr->va_size) ||
filsz > memsz) {
Solve the problem where it is possible to get the kernel stuck in a loop down in pmap_init_pt(). A subtraction causes the number of pages to become negative, that was assigned to an unsigned variable, and there is a lot of iteration. The bug is due to the ELF image activator not properly checking for its files being the correct size as specified by the ELF header. The solution is to check that the header doesn't ask for part of a file when that part of the file doesn't exist. Make sure to set VEXEC at the proper times to make the executables immutable (remove race conditions). Also, the ELF format specifiies header entries that allow embedding of other executables (hence how ld-elf.so.1 gets loaded, but not the same as loading shared libraries), so those executables need to be set VEXEC, too, so they're immutable. Reviewed by: peter
2000-07-23 06:49:46 +00:00
uprintf("elf_load_section: truncated ELF file\n");
return (ENOEXEC);
}
Improve abstraction. Eliminate direct access by elf*_load_section() to an OBJT_VNODE-specific field of the vm object. The same information can be just as easily obtained from the struct vattr that is in struct image_params if the latter is passed to elf*_load_section(). Moreover, by replacing the vmspace and vm object parameters to elf*_load_section() with a struct image_params parameter, we actually reduce the size of the object code. In collaboration with: kib
2012-01-17 00:27:32 +00:00
object = imgp->object;
map = &imgp->proc->p_vmspace->vm_map;
Remove trunc_page_ps() and round_page_ps() macros. This completes the undoing of r100384. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19680
2019-03-23 13:41:14 +00:00
map_addr = trunc_page((vm_offset_t)vmaddr);
file_addr = trunc_page(offset);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
/*
* We have two choices. We can either clear the data in the last page
* of an oversized mapping, or we can start the anon mapping a page
* early and copy the initialized data into that first page. We
Style and punctuation fixes. Reviewed by: kib MFC after: 3 days
2017-03-05 23:59:04 +00:00
* choose the second.
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
*/
Accept linkers representation for ELF segments with zero on-disk length. For such segments, GNU bfd linker writes knowingly incorrect value into the the file offset field of the program header entry, with the motivation that file should not be mapped for creation of this segment at all. Relax checks for the ELF structure validity when on-disk segment length is zero, and explicitely set mapping length to zero for such segments to avoid validating rounding arithmetic. PR: 217610 Reported by: Robert Clausecker <fuz@fuz.su> Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-12 13:51:13 +00:00
if (filsz == 0)
map_len = 0;
else if (memsz > filsz)
Remove trunc_page_ps() and round_page_ps() macros. This completes the undoing of r100384. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19680
2019-03-23 13:41:14 +00:00
map_len = trunc_page(offset + filsz) - file_addr;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
else
Remove trunc_page_ps() and round_page_ps() macros. This completes the undoing of r100384. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19680
2019-03-23 13:41:14 +00:00
map_len = round_page(offset + filsz) - file_addr;
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
if (map_len != 0) {
Change the way ELF coredumps are handled. Instead of unconditionally skipping read-only pages, which can result in valuable non-text-related data not getting dumped, the ELF loader and the dynamic loader now mark read-only text pages NOCORE and the coredump code only checks (primarily) for complete inaccessibility of the page or NOCORE being set. Certain applications which map large amounts of read-only data will produce much larger cores. A new sysctl has been added, debug.elf_legacy_coredump, which will revert to the old behavior. This commit represents collaborative work by all parties involved. The PR contains a program demonstrating the problem. PR: kern/45994 Submitted by: "Peter Edwards" <pmedwards@eircom.net>, Archie Cobbs <archie@dellroad.org> Reviewed by: jdp, dillon MFC after: 7 days
2002-12-16 19:24:43 +00:00
/* cow flags: don't dump readonly sections in core */
cow = MAP_COPY_ON_WRITE | MAP_PREFAULT |
(prot & VM_PROT_WRITE ? 0 : MAP_DISABLE_COREDUMP);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
rv = __elfN(map_insert)(imgp, map, object, file_addr,
map_addr, map_addr + map_len, prot, cow);
Avoid a vm object reference leak in a rarely used code path. An executable contains at most one PT_INTERP program header. Therefore, the loop that searches for it can terminate after it is found rather than iterating over the entire set of program headers. Eliminate an unneeded initialization. Reviewed by: tegge
2006-01-21 20:11:49 +00:00
if (rv != KERN_SUCCESS)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (EINVAL);
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
/* we can stop now if we've covered it all */
Style. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-12 13:49:42 +00:00
if (memsz == filsz)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (0);
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
* We have to get the remaining bit of the file into the first part
* of the oversized map segment. This is normally because the .data
* segment in the file is extended to provide bss. It's a neat idea
* to try and save a page, but it's a pain in the behind to implement.
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
*/
Remove trunc_page_ps() and round_page_ps() macros. This completes the undoing of r100384. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19680
2019-03-23 13:41:14 +00:00
copy_len = filsz == 0 ? 0 : (offset + filsz) - trunc_page(offset +
filsz);
map_addr = trunc_page((vm_offset_t)vmaddr + filsz);
map_len = round_page((vm_offset_t)vmaddr + memsz) - map_addr;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
/* This had damn well better be true! */
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
if (map_len != 0) {
When deallocating the vm object in elf_map_insert() due to vm_map_insert() failure, drop the vnode lock around the call to vm_object_deallocate(). Since the deallocated object is the vm object of the vnode, we might get the vnode lock recursion there. In fact, it is almost impossible to make vm_map_insert() failing there on stock kernel. Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2017-03-01 10:22:07 +00:00
rv = __elfN(map_insert)(imgp, map, NULL, 0, map_addr,
Avoid unnecessary calls to vm_map_protect() in elf_load_section(). Typically, when elf_load_section() unconditionally passed VM_PROT_ALL to elf_map_insert(), it was needlessly enabling execute access on the mapping, and it would later have to call vm_map_protect() to correct the mapping's access rights. Now, instead, elf_load_section() always passes its parameter "prot" to elf_map_insert(). So, elf_load_section() must only call vm_map_protect() if it needs to remove the write access that was temporarily granted to perform a copyout(). Reviewed by: kib MFC after: 1 week
2017-03-18 23:37:00 +00:00
map_addr + map_len, prot, 0);
Style. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-12 13:49:42 +00:00
if (rv != KERN_SUCCESS)
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (EINVAL);
Fix some rounding problems.. In some (fairly rare) situtaions it mapped one page too many, which caused obreak() to fail in vm_map_find() with ENOMEM because of the conflicting page.
1996-03-10 22:37:34 +00:00
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
if (copy_len != 0) {
Use sf_buf_alloc() instead of vm_map_find() on exec_map to create the ephemeral mappings that are used as the source for three copy operations from kernel space to user space. There are two reasons for making this change: (1) Under heavy load exec_map can fill up causing vm_map_find() to fail. When it fails, the nascent process is aborted (SIGABRT). Whereas, this reimplementation using sf_buf_alloc() sleeps. (2) Although it is possible to sleep on vm_map_find()'s failure until address space becomes available (see kmem_alloc_wait()), using sf_buf_alloc() is faster. Furthermore, the reimplementation uses a CPU private mapping, avoiding a TLB shootdown on multiprocessors. Problem uncovered by: kris@ Reviewed by: tegge@ MFC after: 3 weeks
2005-12-16 18:34:14 +00:00
sf = vm_imgact_map_page(object, offset + filsz);
if (sf == NULL)
return (EIO);
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
/* send the page fragment to user space */
Remove a redundant offset computation in elf_load_section(). With r344705 the offset is always zero. Submitted by: Wuyang Chung <wuyang.chung1@gmail.com>
2019-07-24 15:18:05 +00:00
error = copyout((caddr_t)sf_buf_kva(sf), (caddr_t)map_addr,
copy_len);
Revert r253939: We cannot busy a page before doing pagefaults. Infact, it can deadlock against vnode lock, as it tries to vget(). Other functions, right now, have an opposite lock ordering, like vm_object_sync(), which acquires the vnode lock first and then sleeps on the busy mechanism. Before this patch is reinserted we need to break this ordering. Sponsored by: EMC / Isilon storage division Reported by: kib
2013-08-05 08:55:35 +00:00
vm_imgact_unmap_page(sf);
Style. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2017-03-12 13:49:42 +00:00
if (error != 0)
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
return (error);
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
Avoid unnecessary calls to vm_map_protect() in elf_load_section(). Typically, when elf_load_section() unconditionally passed VM_PROT_ALL to elf_map_insert(), it was needlessly enabling execute access on the mapping, and it would later have to call vm_map_protect() to correct the mapping's access rights. Now, instead, elf_load_section() always passes its parameter "prot" to elf_map_insert(). So, elf_load_section() must only call vm_map_protect() if it needs to remove the write access that was temporarily granted to perform a copyout(). Reviewed by: kib MFC after: 1 week
2017-03-18 23:37:00 +00:00
* Remove write access to the page if it was only granted by map_insert
* to allow copyout.
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
*/
Avoid unnecessary calls to vm_map_protect() in elf_load_section(). Typically, when elf_load_section() unconditionally passed VM_PROT_ALL to elf_map_insert(), it was needlessly enabling execute access on the mapping, and it would later have to call vm_map_protect() to correct the mapping's access rights. Now, instead, elf_load_section() always passes its parameter "prot" to elf_map_insert(). So, elf_load_section() must only call vm_map_protect() if it needs to remove the write access that was temporarily granted to perform a copyout(). Reviewed by: kib MFC after: 1 week
2017-03-18 23:37:00 +00:00
if ((prot & VM_PROT_WRITE) == 0)
vm_map_protect(map, trunc_page(map_addr), round_page(map_addr +
map_len), prot, FALSE);
Fix some rounding problems.. In some (fairly rare) situtaions it mapped one page too many, which caused obreak() to fail in vm_map_find() with ENOMEM because of the conflicting page.
1996-03-10 22:37:34 +00:00
Eliminate an unneeded (vm_prot_t) parameter from two functions. Eliminate unnecessary uses of a local variable. Reviewed by: tegge
2005-12-20 23:42:18 +00:00
return (0);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
static int
__elfN(load_sections)(struct image_params *imgp, const Elf_Ehdr *hdr,
const Elf_Phdr *phdr, u_long rbase, u_long *base_addrp)
{
vm_prot_t prot;
u_long base_addr;
bool first;
int error, i;
Improve vnode lock assertions. MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-04-10 10:21:14 +00:00
ASSERT_VOP_LOCKED(imgp->vp, __func__);
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
base_addr = 0;
first = true;
for (i = 0; i < hdr->e_phnum; i++) {
if (phdr[i].p_type != PT_LOAD || phdr[i].p_memsz == 0)
continue;
/* Loadable segment */
prot = __elfN(trans_prot)(phdr[i].p_flags);
error = __elfN(load_section)(imgp, phdr[i].p_offset,
(caddr_t)(uintptr_t)phdr[i].p_vaddr + rbase,
phdr[i].p_memsz, phdr[i].p_filesz, prot);
if (error != 0)
return (error);
/*
* Establish the base address if this is the first segment.
*/
if (first) {
base_addr = trunc_page(phdr[i].p_vaddr + rbase);
first = false;
}
}
if (base_addrp != NULL)
*base_addrp = base_addr;
return (0);
}
If you merge this into -stable, please increment __FreeBSD_version in "src/sys/sys/param.h". Fix the ELF image activator so that it can handle dynamic linkers which are executables linked at a fixed address. This improves compliance with the ABI spec, and it opens the door to possibly better dynamic linker performance in the future. I've experimented a bit with a fixed-address dynamic linker, and it works fine. But I don't have any measurements yet to determine whether it's worthwhile. Also, remove a few calculations that were never used for anything. I will increment __FreeBSD_version, since this adds a new capability to the kernel that the dynamic linker might some day rely upon.
1999-02-20 23:52:34 +00:00
/*
* Load the file "file" into memory. It may be either a shared object
* or an executable.
*
* The "addr" reference parameter is in/out. On entry, it specifies
* the address where a shared object should be loaded. If the file is
* an executable, this value is ignored. On exit, "addr" specifies
* where the file was actually loaded.
*
* The "entry" reference parameter is out only. On exit, it specifies
* the entry point for the loaded file.
*/
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
static int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__elfN(load_file)(struct proc *p, const char *file, u_long *addr,
Remove sv_pagesize, originally introduced with r100384. In all of the architectures we have today, we always use PAGE_SIZE. While in theory one could define different things, none of the current architectures do, even the ones that have transitioned from 32-bit to 64-bit like i386 and arm. Some ancient mips binaries on other systems used 8k instead of 4k, but we don't support running those and likely never will due to their age and obscurity. Reviewed by: imp (who also contributed the commit message) Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19280
2019-03-01 16:16:38 +00:00
u_long *entry)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
struct {
struct nameidata nd;
struct vattr attr;
struct image_params image_params;
} *tempdata;
Fix warnings in preparation for adding -Wall -Wcast-qual to the kernel compile
1999-01-27 21:50:00 +00:00
const Elf_Ehdr *hdr = NULL;
const Elf_Phdr *phdr = NULL;
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
struct nameidata *nd;
struct vattr *attr;
struct image_params *imgp;
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
u_long rbase;
If you merge this into -stable, please increment __FreeBSD_version in "src/sys/sys/param.h". Fix the ELF image activator so that it can handle dynamic linkers which are executables linked at a fixed address. This improves compliance with the ABI spec, and it opens the door to possibly better dynamic linker performance in the future. I've experimented a bit with a fixed-address dynamic linker, and it works fine. But I don't have any measurements yet to determine whether it's worthwhile. Also, remove a few calculations that were never used for anything. I will increment __FreeBSD_version, since this adds a new capability to the kernel that the dynamic linker might some day rely upon.
1999-02-20 23:52:34 +00:00
u_long base_addr = 0;
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
int error;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Add some checks to ensure that Capsicum is behaving correctly, and add some more explicit comments about what's going on and what future maintainers need to do when e.g. adding a new operation to a sys_machdep.c. Approved by: mentor(rwatson), re(bz)
2011-06-30 10:56:02 +00:00
#ifdef CAPABILITY_MODE
/*
* XXXJA: This check can go away once we are sufficiently confident
* that the checks in namei() are correct.
*/
if (IN_CAPABILITY_MODE(curthread))
return (ECAPMODE);
#endif
When loading ELF interpreter, initialize whole nested image_params with zero. Otherwise we could mishandle imgp->textset. Reviewed by: markj MFC after: 1 week Differential revision: https://reviews.freebsd.org/D21560
2019-09-07 16:03:26 +00:00
tempdata = malloc(sizeof(*tempdata), M_TEMP, M_WAITOK | M_ZERO);
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
nd = &tempdata->nd;
attr = &tempdata->attr;
imgp = &tempdata->image_params;
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
/*
* Initialize part of the common data
*/
imgp->proc = p;
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
imgp->attr = attr;
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
Set ISOPEN in namei flags when opening executable interpreters. These vnodes are explicitly opened via VOP_OPEN via exec_check_permissions identical to the main exectuable image. Setting ISOPEN allows filesystems to perform suitable checks in VOP_LOOKUP (e.g. close-to-open consistency in the NFS client). Reviewed by: kib MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D21129
2019-08-03 01:02:52 +00:00
NDINIT(nd, LOOKUP, ISOPEN | FOLLOW | LOCKSHARED | LOCKLEAF,
UIO_SYSSPACE, file, curthread);
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
if ((error = namei(nd)) != 0) {
nd->ni_vp = NULL;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
goto fail;
}
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
NDFREE(nd, NDF_ONLY_PNBUF);
imgp->vp = nd->ni_vp;
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
* Check permissions, modes, uid, etc on the file, and "open" it.
*/
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
error = exec_check_permissions(imgp);
Maintain the vnode lock throughout elfN_load_file() rather than releasing it and reacquiring it in vrele(). Consequently, there is no reason to increase the reference count on the vm object caching the file's pages. Reviewed by: tegge Eliminate unused parameters to elfN_load_file().
2005-12-21 18:58:40 +00:00
if (error)
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
goto fail;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
error = exec_map_first_page(imgp);
Maintain the vnode lock throughout elfN_load_file() rather than releasing it and reacquiring it in vrele(). Consequently, there is no reason to increase the reference count on the vm object caching the file's pages. Reviewed by: tegge Eliminate unused parameters to elfN_load_file().
2005-12-21 18:58:40 +00:00
if (error)
goto fail;
Don't use VOP_GETVOBJECT, use vp->v_object directly.
2005-01-25 00:40:01 +00:00
imgp->object = nd->ni_vp->v_object;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Fix warnings in preparation for adding -Wall -Wcast-qual to the kernel compile
1999-01-27 21:50:00 +00:00
hdr = (const Elf_Ehdr *)imgp->image_header;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
if ((error = __elfN(check_header)(hdr)) != 0)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
goto fail;
If you merge this into -stable, please increment __FreeBSD_version in "src/sys/sys/param.h". Fix the ELF image activator so that it can handle dynamic linkers which are executables linked at a fixed address. This improves compliance with the ABI spec, and it opens the door to possibly better dynamic linker performance in the future. I've experimented a bit with a fixed-address dynamic linker, and it works fine. But I don't have any measurements yet to determine whether it's worthwhile. Also, remove a few calculations that were never used for anything. I will increment __FreeBSD_version, since this adds a new capability to the kernel that the dynamic linker might some day rely upon.
1999-02-20 23:52:34 +00:00
if (hdr->e_type == ET_DYN)
rbase = *addr;
else if (hdr->e_type == ET_EXEC)
rbase = 0;
else {
error = ENOEXEC;
goto fail;
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Verify more bits of the ELF header: the program header table entry size and the ELF version. Also, avoid a potential integer overflow when determining whether the ELF header fits entirely within the first page. Reviewed by: jdp A panic when attempting to execute an ELF binary with a bogus program header table entry size was Reported by: Christer Öberg <christer.oberg@texonet.com>
2004-03-18 16:33:05 +00:00
/* Only support headers that fit within first page for now */
Some cleanups and optimizations: - Use the system headers method for Elf32/Elf64 symbol compatability - get rid of the UPRINTF debugging. - check the ELF header for compatability much more completely - optimize the section mapper. Use the same direct VM interfaces that imgact_aout.c and kern_exec.c use. - Check the return codes from the vm_* functions better. Some return KERN_* results, not an errno. - prefault the page tables to reduce startup faults on page tables like a.out does. - reset the segment protection to zero for each loop, otherwise each segment could get progressively more privs. (eg: if the first was read/write/execute, and the second was meant to be read/execute, the bug would make the second r/w/x too. In practice this was not a problem because executables are normally laid out with text first.) - Don't impose arbitary limits. Use the limits on headers imposed by the need to fit them into one page. - Remove unused switch() cases now that the verbose debugging is gone. I've been using an earlier version of this for a month or so. This sped up ELF exec speed a bit for me but I found it hard to get consistant benchmarks when I tested it last (a few weeks ago). I'm still bothered by the page read out of order caused by the transition from data to bss. This which requires either part filling the transition page or clearing the remainder.
1998-10-18 15:55:12 +00:00
if ((hdr->e_phoff > PAGE_SIZE) ||
- Fix two possible overflows when testing if ELF program headers are on the first page: 1. Cast uint16_t operands in a multiplication to unsigned int because otherwise the implicit promotion to int results in a signed multiplication that can overflow and the behaviour on integer overflow is undefined. 2. Replace (offset + size > PAGE_SIZE) with (size > PAGE_SIZE - offset) because the sum may overflow. - Use the same tests to see if the path to the interpreter is on the first page. There's no overflow here because size is already limited by MAXPATHLEN, but the compiler optimises the new tests better. Also fix an off-by-one error. - Simplify tests to see if an ELF note program header is on the first page. This also fixes an off-by-one error. Reviewed by: kib MFC after: 1 week
2013-03-13 22:01:31 +00:00
(u_int)hdr->e_phentsize * hdr->e_phnum > PAGE_SIZE - hdr->e_phoff) {
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
error = ENOEXEC;
goto fail;
}
Fix warnings in preparation for adding -Wall -Wcast-qual to the kernel compile
1999-01-27 21:50:00 +00:00
phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff);
Check for the program headers alignment of the ELF images before dereferencing. Unaligned access could cause panic on strict alignment architectures. Reviewed by: marcel, marius (also tested on sparc64, thanks !) MFC after: 3 days
2007-12-04 12:21:27 +00:00
if (!aligned(phdr, Elf_Addr)) {
error = ENOEXEC;
goto fail;
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
error = __elfN(load_sections)(imgp, hdr, phdr, rbase, &base_addr);
if (error != 0)
goto fail;
If you merge this into -stable, please increment __FreeBSD_version in "src/sys/sys/param.h". Fix the ELF image activator so that it can handle dynamic linkers which are executables linked at a fixed address. This improves compliance with the ABI spec, and it opens the door to possibly better dynamic linker performance in the future. I've experimented a bit with a fixed-address dynamic linker, and it works fine. But I don't have any measurements yet to determine whether it's worthwhile. Also, remove a few calculations that were never used for anything. I will increment __FreeBSD_version, since this adds a new capability to the kernel that the dynamic linker might some day rely upon.
1999-02-20 23:52:34 +00:00
*addr = base_addr;
Fixed white space around operators, casts and reserved words. Reviewed by: md5
2002-08-24 22:55:16 +00:00
*entry = (unsigned long)hdr->e_entry + rbase;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
fail:
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
if (imgp->firstpage)
exec_unmap_first_page(imgp);
Clean up execve locking: - Grab the vnode object early in exec when we still have the vnode lock. - Cache the object in the image_params. - Make use of the cached object in imgact_*.c
2002-07-06 07:00:01 +00:00
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
if (nd->ni_vp) {
if (imgp->textset)
VOP_UNSET_TEXT_CHECKED(nd->ni_vp);
Maintain the vnode lock throughout elfN_load_file() rather than releasing it and reacquiring it in vrele(). Consequently, there is no reason to increase the reference count on the vm object caching the file's pages. Reviewed by: tegge Eliminate unused parameters to elfN_load_file().
2005-12-21 18:58:40 +00:00
vput(nd->ni_vp);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
}
Reduce stack allocation (stack-fast?). elf_load_file() => 352 to 52 bytes exec_elf_imgact() => 1072 to 48 bytes elf_corehdr() => 396 to 8 bytes Reviewed by: julian
2001-08-16 16:14:26 +00:00
free(tempdata, M_TEMP);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (error);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
static u_long
__CONCAT(rnd_, __elfN(base))(vm_map_t map __unused, u_long minv, u_long maxv,
u_int align)
{
u_long rbase, res;
MPASS(vm_map_min(map) <= minv);
MPASS(maxv <= vm_map_max(map));
MPASS(minv < maxv);
MPASS(minv + align < maxv);
arc4rand(&rbase, sizeof(rbase), 0);
res = roundup(minv, (u_long)align) + rbase % (maxv - minv);
res &= ~((u_long)align - 1);
if (res >= maxv)
res -= align;
KASSERT(res >= minv,
("res %#lx < minv %#lx, maxv %#lx rbase %#lx",
res, minv, maxv, rbase));
KASSERT(res < maxv,
("res %#lx > maxv %#lx, minv %#lx rbase %#lx",
res, maxv, minv, rbase));
return (res);
}
Factor out resource limit enforcement code in the ELF loader. It makes the code slightly easier to follow, and might make it easier to fix the resouce accounting to also account for the interpreter. The PROC_UNLOCK() is moved earlier - I don't see anything it should protect; the lim_max() is a wrapper around lim_rlimit(), and that, differently from lim_rlimit_proc(), doesn't require the proc lock to be held. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19689
2019-03-26 15:35:49 +00:00
static int
__elfN(enforce_limits)(struct image_params *imgp, const Elf_Ehdr *hdr,
const Elf_Phdr *phdr, u_long et_dyn_addr)
{
struct vmspace *vmspace;
const char *err_str;
u_long text_size, data_size, total_size, text_addr, data_addr;
u_long seg_size, seg_addr;
int i;
err_str = NULL;
text_size = data_size = total_size = text_addr = data_addr = 0;
for (i = 0; i < hdr->e_phnum; i++) {
if (phdr[i].p_type != PT_LOAD || phdr[i].p_memsz == 0)
continue;
seg_addr = trunc_page(phdr[i].p_vaddr + et_dyn_addr);
seg_size = round_page(phdr[i].p_memsz +
phdr[i].p_vaddr + et_dyn_addr - seg_addr);
/*
* Make the largest executable segment the official
* text segment and all others data.
*
* Note that obreak() assumes that data_addr + data_size == end
* of data load area, and the ELF file format expects segments
* to be sorted by address. If multiple data segments exist,
* the last one will be used.
*/
if ((phdr[i].p_flags & PF_X) != 0 && text_size < seg_size) {
text_size = seg_size;
text_addr = seg_addr;
} else {
data_size = seg_size;
data_addr = seg_addr;
}
total_size += seg_size;
}
if (data_addr == 0 && data_size == 0) {
data_addr = text_addr;
data_size = text_size;
}
/*
* Check limits. It should be safe to check the
* limits after loading the segments since we do
* not actually fault in all the segments pages.
*/
PROC_LOCK(imgp->proc);
if (data_size > lim_cur_proc(imgp->proc, RLIMIT_DATA))
err_str = "Data segment size exceeds process limit";
else if (text_size > maxtsiz)
err_str = "Text segment size exceeds system limit";
else if (total_size > lim_cur_proc(imgp->proc, RLIMIT_VMEM))
err_str = "Total segment size exceeds process limit";
else if (racct_set(imgp->proc, RACCT_DATA, data_size) != 0)
err_str = "Data segment size exceeds resource limit";
else if (racct_set(imgp->proc, RACCT_VMEM, total_size) != 0)
err_str = "Total segment size exceeds resource limit";
PROC_UNLOCK(imgp->proc);
if (err_str != NULL) {
uprintf("%s\n", err_str);
return (ENOMEM);
}
vmspace = imgp->proc->p_vmspace;
vmspace->vm_tsize = text_size >> PAGE_SHIFT;
vmspace->vm_taddr = (caddr_t)(uintptr_t)text_addr;
vmspace->vm_dsize = data_size >> PAGE_SHIFT;
vmspace->vm_daddr = (caddr_t)(uintptr_t)data_addr;
return (0);
}
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
static int
__elfN(get_interp)(struct image_params *imgp, const Elf_Phdr *phdr,
char **interpp, bool *free_interpp)
{
struct thread *td;
char *interp;
int error, interp_name_len;
KASSERT(phdr->p_type == PT_INTERP,
("%s: p_type %u != PT_INTERP", __func__, phdr->p_type));
Improve vnode lock assertions. MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-04-10 10:21:14 +00:00
ASSERT_VOP_LOCKED(imgp->vp, __func__);
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
td = curthread;
/* Path to interpreter */
if (phdr->p_filesz < 2 || phdr->p_filesz > MAXPATHLEN) {
uprintf("Invalid PT_INTERP\n");
return (ENOEXEC);
}
interp_name_len = phdr->p_filesz;
if (phdr->p_offset > PAGE_SIZE ||
interp_name_len > PAGE_SIZE - phdr->p_offset) {
imgact_elf.c: Add comment explaining the malloc/VOP_UNLOCK() dance from r347148. Requested by: alc Sponsored by: The FreeBSD Foundation MFC after: 3 days
2019-05-16 13:03:54 +00:00
/*
Grammar fixes for r347690. Submitted by: alc MFC after: 3 days
2019-05-17 21:18:11 +00:00
* The vnode lock might be needed by the pagedaemon to
imgact_elf.c: Add comment explaining the malloc/VOP_UNLOCK() dance from r347148. Requested by: alc Sponsored by: The FreeBSD Foundation MFC after: 3 days
2019-05-16 13:03:54 +00:00
* clean pages owned by the vnode. Do not allow sleep
* waiting for memory with the vnode locked, instead
* try non-sleepable allocation first, and if it
* fails, go to the slow path were we drop the lock
Grammar fixes for r347690. Submitted by: alc MFC after: 3 days
2019-05-17 21:18:11 +00:00
* and do M_WAITOK. A text reference prevents
* modifications to the vnode content.
imgact_elf.c: Add comment explaining the malloc/VOP_UNLOCK() dance from r347148. Requested by: alc Sponsored by: The FreeBSD Foundation MFC after: 3 days
2019-05-16 13:03:54 +00:00
*/
imgact_elf: do not relock the text vnode if possible. We unlock the vnode around malloc(M_WAITOK), to make it possible for pagedaemon to flush vnode pages for us. Instead of doing it unconditionally, first try M_NOWAIT allocation, which typically succeed. Only on failure, unlock the vnode and retry with M_WAITOK. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:04:01 +00:00
interp = malloc(interp_name_len + 1, M_TEMP, M_NOWAIT);
if (interp == NULL) {
VOP_UNLOCK(imgp->vp, 0);
interp = malloc(interp_name_len + 1, M_TEMP, M_WAITOK);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
imgact_elf: do not relock the text vnode if possible. We unlock the vnode around malloc(M_WAITOK), to make it possible for pagedaemon to flush vnode pages for us. Instead of doing it unconditionally, first try M_NOWAIT allocation, which typically succeed. Only on failure, unlock the vnode and retry with M_WAITOK. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:04:01 +00:00
}
imgact_elf.c: Add comment explaining the malloc/VOP_UNLOCK() dance from r347148. Requested by: alc Sponsored by: The FreeBSD Foundation MFC after: 3 days
2019-05-16 13:03:54 +00:00
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
error = vn_rdwr(UIO_READ, imgp->vp, interp,
interp_name_len, phdr->p_offset,
UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
NOCRED, NULL, td);
if (error != 0) {
free(interp, M_TEMP);
uprintf("i/o error PT_INTERP %d\n", error);
return (error);
}
interp[interp_name_len] = '\0';
*interpp = interp;
*free_interpp = true;
return (0);
}
interp = __DECONST(char *, imgp->image_header) + phdr->p_offset;
if (interp[interp_name_len - 1] != '\0') {
uprintf("Invalid PT_INTERP\n");
return (ENOEXEC);
}
*interpp = interp;
*free_interpp = false;
return (0);
}
Refactor ELF interpreter loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19741
2019-04-08 14:31:07 +00:00
static int
__elfN(load_interp)(struct image_params *imgp, const Elf_Brandinfo *brand_info,
const char *interp, u_long *addr, u_long *entry)
{
char *path;
int error;
if (brand_info->emul_path != NULL &&
brand_info->emul_path[0] != '\0') {
path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
snprintf(path, MAXPATHLEN, "%s%s",
brand_info->emul_path, interp);
error = __elfN(load_file)(imgp->proc, path, addr, entry);
free(path, M_TEMP);
if (error == 0)
return (0);
}
if (brand_info->interp_newpath != NULL &&
(brand_info->interp_path == NULL ||
strcmp(interp, brand_info->interp_path) == 0)) {
error = __elfN(load_file)(imgp->proc,
brand_info->interp_newpath, addr, entry);
if (error == 0)
return (0);
}
error = __elfN(load_file)(imgp->proc, interp, addr, entry);
if (error == 0)
return (0);
uprintf("ELF interpreter %s not found, error %d\n", interp, error);
return (error);
}
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
/*
* Impossible et_dyn_addr initial value indicating that the real base
* must be calculated later with some randomization applied.
*/
#define ET_DYN_ADDR_RAND 1
Staticize.
1998-02-09 06:11:36 +00:00
static int
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
__CONCAT(exec_, __elfN(imgact))(struct image_params *imgp)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
struct thread *td;
const Elf_Ehdr *hdr;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
const Elf_Phdr *phdr;
Avoid a vm object reference leak in a rarely used code path. An executable contains at most one PT_INTERP program header. Therefore, the loop that searches for it can terminate after it is found rather than iterating over the entire set of program headers. Eliminate an unneeded initialization. Reviewed by: tegge
2006-01-21 20:11:49 +00:00
Elf_Auxargs *elf_auxargs;
Fully implement vfork. Vfork is now much much faster than even our fork. (On my machine, fork is about 240usecs, vfork is 78usecs.) Implement rfork(!RFPROC !RFMEM), which allows a thread to divorce its memory from the other threads of a group. Implement rfork(!RFPROC RFCFDG), which closes all file descriptors, eliminating possible existing shares with other threads/processes. Implement rfork(!RFPROC RFFDG), which divorces the file descriptors for a thread from the rest of the group. Fix the case where a thread does an exec. It is almost nonsense for a thread to modify the other threads address space by an exec, so we now automatically divorce the address space before modifying it.
1997-04-13 01:48:35 +00:00
struct vmspace *vmspace;
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
vm_map_t map;
Refactor ELF interpreter loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19741
2019-04-08 14:31:07 +00:00
char *interp;
If an ELF executable has a recognized brand, then believe it. Formerly, the heuristic involving the interpreter path took precedence. Also, print a better error message if the brand is missing or not recognized. If there is no brand at all, give the user a hint that "brandelf" needs to be run.
1998-10-11 19:22:07 +00:00
Elf_Brandinfo *brand_info;
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
struct sysentvec *sv;
Factor out resource limit enforcement code in the ELF loader. It makes the code slightly easier to follow, and might make it easier to fix the resouce accounting to also account for the interpreter. The PROC_UNLOCK() is moved earlier - I don't see anything it should protect; the lim_max() is a wrapper around lim_rlimit(), and that, differently from lim_rlimit_proc(), doesn't require the proc lock to be held. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19689
2019-03-26 15:35:49 +00:00
u_long addr, baddr, et_dyn_addr, entry, proghdr;
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
u_long maxalign, mapsz, maxv, maxv1;
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
uint32_t fctl0;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
int32_t osrel;
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
bool free_interp;
Refactor ELF interpreter loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19741
2019-04-08 14:31:07 +00:00
int error, i, n;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
hdr = (const Elf_Ehdr *)imgp->image_header;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
* Do we have a valid ELF header ?
Remove kern.elf32.can_exec_dyn sysctl. Instead extend Brandinfo structure with flags bitfield and set BI_CAN_EXEC_DYN flag for all brands that usually allow executing elf dynamic binaries (aka shared libraries). When it is requested to execute ET_DYN elf image check if this flag is on after we know the elf brand allowing execution if so. PR: kern/87615 Submitted by: Marcin Koziej <creep@desk.pl>
2005-12-26 21:23:57 +00:00
*
* Only allow ET_EXEC & ET_DYN here, reject ET_DYN later
* if particular brand doesn't support it.
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
*/
Remove kern.elf32.can_exec_dyn sysctl. Instead extend Brandinfo structure with flags bitfield and set BI_CAN_EXEC_DYN flag for all brands that usually allow executing elf dynamic binaries (aka shared libraries). When it is requested to execute ET_DYN elf image check if this flag is on after we know the elf brand allowing execution if so. PR: kern/87615 Submitted by: Marcin Koziej <creep@desk.pl>
2005-12-26 21:23:57 +00:00
if (__elfN(check_header)(hdr) != 0 ||
(hdr->e_type != ET_EXEC && hdr->e_type != ET_DYN))
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (-1);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
* From here on down, we return an errno, not -1, as we've
* detected an ELF file.
*/
if ((hdr->e_phoff > PAGE_SIZE) ||
- Fix two possible overflows when testing if ELF program headers are on the first page: 1. Cast uint16_t operands in a multiplication to unsigned int because otherwise the implicit promotion to int results in a signed multiplication that can overflow and the behaviour on integer overflow is undefined. 2. Replace (offset + size > PAGE_SIZE) with (size > PAGE_SIZE - offset) because the sum may overflow. - Use the same tests to see if the path to the interpreter is on the first page. There's no overflow here because size is already limited by MAXPATHLEN, but the compiler optimises the new tests better. Also fix an off-by-one error. - Simplify tests to see if an ELF note program header is on the first page. This also fixes an off-by-one error. Reviewed by: kib MFC after: 1 week
2013-03-13 22:01:31 +00:00
(u_int)hdr->e_phentsize * hdr->e_phnum > PAGE_SIZE - hdr->e_phoff) {
Update the ELF image activator to use some of the exec resources rather than rolling it's own. This means that it now uses the "safe" exec_map_first_page() to get the ld.so headers rather than risking a panic on a page fault failure (eg: NFS server goes down). Since all the ELF tools go to a lot of trouble to make sure everything lives in the first page for executables, this is a win. I have not seen any ELF executable on any system where all the headers didn't fit in the first page with lots of room to spare. I have been running variations of this code for some time on my pure ELF systems.
1998-03-02 05:47:58 +00:00
/* Only support headers in first page for now */
Add user facing errors for exceeding process memory limits Previously the process terminating with SIGABRT at startup was the only notification. PR: 200617 Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2731
2015-06-08 16:07:07 +00:00
uprintf("Program headers not in the first page\n");
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (ENOEXEC);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
Add user facing errors for exceeding process memory limits Previously the process terminating with SIGABRT at startup was the only notification. PR: 200617 Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2731
2015-06-08 16:07:07 +00:00
phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff);
if (!aligned(phdr, Elf_Addr)) {
uprintf("Unaligned program headers\n");
Check for the program headers alignment of the ELF images before dereferencing. Unaligned access could cause panic on strict alignment architectures. Reviewed by: marcel, marius (also tested on sparc64, thanks !) MFC after: 3 days
2007-12-04 12:21:27 +00:00
return (ENOEXEC);
Add user facing errors for exceeding process memory limits Previously the process terminating with SIGABRT at startup was the only notification. PR: 200617 Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2731
2015-06-08 16:07:07 +00:00
}
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
n = error = 0;
If ET_DYN binary has non-zero base address for some reason, honour it and do not relocate the binary to ET_DYN_LOAD_ADDR. This allows for the binary author to influence address map of the process. In particular, when the binary is actually an interpeter, this allows to have almost usual process address map. Communicate the relocation bias of the mapping for interpeter-less ET_DYN binary, that is interperter itself, in AT_BASE aux entry. This way, rtld is able to find its dynamic structure and relocate itself. Note that mapbase in the rtld is still wrong and requires further fixing. Reported and tested by: rwatson Discussed with: kan MFC after: 3 days
2009-10-18 12:57:48 +00:00
baddr = 0;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
osrel = 0;
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
fctl0 = 0;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
entry = proghdr = 0;
Refactor ELF interpreter loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19741
2019-04-08 14:31:07 +00:00
interp = NULL;
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
free_interp = false;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
td = curthread;
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
maxalign = PAGE_SIZE;
mapsz = 0;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
for (i = 0; i < hdr->e_phnum; i++) {
In elf image activator, read and apply the stack protection mode from PT_GNU_STACK program header, if present and enabled. Two new sysctls are provided, kern.elf32.nxstack and kern.elf64.nxstack, that allow to enable PT_GNU_STACK for ABIs of specified bitsize, if ABI decided to support shared page. Inform rtld about access mode of the stack initial mapping by AT_STACKPROT aux vector. At the moment, the default is disabled, waiting for the usermode support bits.
2011-01-08 16:30:59 +00:00
switch (phdr[i].p_type) {
case PT_LOAD:
If ET_DYN binary has non-zero base address for some reason, honour it and do not relocate the binary to ET_DYN_LOAD_ADDR. This allows for the binary author to influence address map of the process. In particular, when the binary is actually an interpeter, this allows to have almost usual process address map. Communicate the relocation bias of the mapping for interpeter-less ET_DYN binary, that is interperter itself, in AT_BASE aux entry. This way, rtld is able to find its dynamic structure and relocate itself. Note that mapbase in the rtld is still wrong and requires further fixing. Reported and tested by: rwatson Discussed with: kan MFC after: 3 days
2009-10-18 12:57:48 +00:00
if (n == 0)
baddr = phdr[i].p_vaddr;
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
if (phdr[i].p_align > maxalign)
maxalign = phdr[i].p_align;
mapsz += phdr[i].p_memsz;
If ET_DYN binary has non-zero base address for some reason, honour it and do not relocate the binary to ET_DYN_LOAD_ADDR. This allows for the binary author to influence address map of the process. In particular, when the binary is actually an interpeter, this allows to have almost usual process address map. Communicate the relocation bias of the mapping for interpeter-less ET_DYN binary, that is interperter itself, in AT_BASE aux entry. This way, rtld is able to find its dynamic structure and relocate itself. Note that mapbase in the rtld is still wrong and requires further fixing. Reported and tested by: rwatson Discussed with: kan MFC after: 3 days
2009-10-18 12:57:48 +00:00
n++;
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
/*
* If this segment contains the program headers,
* remember their virtual address for the AT_PHDR
* aux entry. Static binaries don't usually include
* a PT_PHDR entry.
*/
if (phdr[i].p_offset == 0 &&
hdr->e_phoff + hdr->e_phnum * hdr->e_phentsize
<= phdr[i].p_filesz)
proghdr = phdr[i].p_vaddr + hdr->e_phoff;
In elf image activator, read and apply the stack protection mode from PT_GNU_STACK program header, if present and enabled. Two new sysctls are provided, kern.elf32.nxstack and kern.elf64.nxstack, that allow to enable PT_GNU_STACK for ABIs of specified bitsize, if ABI decided to support shared page. Inform rtld about access mode of the stack initial mapping by AT_STACKPROT aux vector. At the moment, the default is disabled, waiting for the usermode support bits.
2011-01-08 16:30:59 +00:00
break;
case PT_INTERP:
Avoid a vm object reference leak in a rarely used code path. An executable contains at most one PT_INTERP program header. Therefore, the loop that searches for it can terminate after it is found rather than iterating over the entire set of program headers. Eliminate an unneeded initialization. Reviewed by: tegge
2006-01-21 20:11:49 +00:00
/* Path to interpreter */
Only allow one PT_INTERP ELF program header. This also fixes a potential memory leak for interp_buf. Differential Revision: https://reviews.freebsd.org/D4692 Reviewed by: kib MFC after: 2 weeks Sponsored by: Juniper Networks
2015-12-24 00:58:11 +00:00
if (interp != NULL) {
uprintf("Multiple PT_INTERP headers\n");
error = ENOEXEC;
goto ret;
}
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
error = __elfN(get_interp)(imgp, &phdr[i], &interp,
&free_interp);
if (error != 0)
goto ret;
In elf image activator, read and apply the stack protection mode from PT_GNU_STACK program header, if present and enabled. Two new sysctls are provided, kern.elf32.nxstack and kern.elf64.nxstack, that allow to enable PT_GNU_STACK for ABIs of specified bitsize, if ABI decided to support shared page. Inform rtld about access mode of the stack initial mapping by AT_STACKPROT aux vector. At the moment, the default is disabled, waiting for the usermode support bits.
2011-01-08 16:30:59 +00:00
break;
case PT_GNU_STACK:
if (__elfN(nxstack))
imgp->stack_prot =
__elfN(trans_prot)(phdr[i].p_flags);
Implement support for binary to requesting specific stack size for the initial thread. It is read by the ELF image activator as the virtual size of the PT_GNU_STACK program header entry, and can be specified by the linker option -z stack-size in newer binutils. The soft RLIMIT_STACK is auto-increased if possible, to satisfy the binary' request. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-04-15 08:13:53 +00:00
imgp->stack_sz = phdr[i].p_memsz;
In elf image activator, read and apply the stack protection mode from PT_GNU_STACK program header, if present and enabled. Two new sysctls are provided, kern.elf32.nxstack and kern.elf64.nxstack, that allow to enable PT_GNU_STACK for ABIs of specified bitsize, if ABI decided to support shared page. Inform rtld about access mode of the stack initial mapping by AT_STACKPROT aux vector. At the moment, the default is disabled, waiting for the usermode support bits.
2011-01-08 16:30:59 +00:00
break;
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
case PT_PHDR: /* Program header table info */
proghdr = phdr[i].p_vaddr;
break;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
}
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
brand_info = __elfN(get_brandinfo)(imgp, interp, &osrel, &fctl0);
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
if (brand_info == NULL) {
uprintf("ELF binary type \"%u\" not known.\n",
hdr->e_ident[EI_OSABI]);
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
error = ENOEXEC;
goto ret;
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
sv = brand_info->sysvec;
imgact_elf: refactor et_dyn_addr calculation This simplifies the logic somewhat. It is extracted from the change in review in D5603. Differential Revision: https://reviews.freebsd.org/D9321
2017-01-24 22:46:43 +00:00
et_dyn_addr = 0;
Map PIE binaries at non-zero base address. Discussed with: bz Reviewed by: kan Tested by: bz (i386, amd64), bsam (linux) MFC after: some time
2009-10-10 15:33:01 +00:00
if (hdr->e_type == ET_DYN) {
Add user facing errors for exceeding process memory limits Previously the process terminating with SIGABRT at startup was the only notification. PR: 200617 Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2731
2015-06-08 16:07:07 +00:00
if ((brand_info->flags & BI_CAN_EXEC_DYN) == 0) {
uprintf("Cannot execute shared object\n");
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
error = ENOEXEC;
goto ret;
Add user facing errors for exceeding process memory limits Previously the process terminating with SIGABRT at startup was the only notification. PR: 200617 Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D2731
2015-06-08 16:07:07 +00:00
}
If ET_DYN binary has non-zero base address for some reason, honour it and do not relocate the binary to ET_DYN_LOAD_ADDR. This allows for the binary author to influence address map of the process. In particular, when the binary is actually an interpeter, this allows to have almost usual process address map. Communicate the relocation bias of the mapping for interpeter-less ET_DYN binary, that is interperter itself, in AT_BASE aux entry. This way, rtld is able to find its dynamic structure and relocate itself. Note that mapbase in the rtld is still wrong and requires further fixing. Reported and tested by: rwatson Discussed with: kan MFC after: 3 days
2009-10-18 12:57:48 +00:00
/*
* Honour the base load address from the dso if it is
* non-zero for some reason.
*/
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
if (baddr == 0) {
if ((sv->sv_flags & SV_ASLR) == 0 ||
(fctl0 & NT_FREEBSD_FCTL_ASLR_DISABLE) != 0)
Make non-ASLR pie base tunable. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-09-21 18:00:23 +00:00
et_dyn_addr = __elfN(pie_base);
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
else if ((__elfN(pie_aslr_enabled) &&
(imgp->proc->p_flag2 & P2_ASLR_DISABLE) == 0) ||
(imgp->proc->p_flag2 & P2_ASLR_ENABLE) != 0)
et_dyn_addr = ET_DYN_ADDR_RAND;
else
Make non-ASLR pie base tunable. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-09-21 18:00:23 +00:00
et_dyn_addr = __elfN(pie_base);
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
}
imgact_elf: refactor et_dyn_addr calculation This simplifies the logic somewhat. It is extracted from the change in review in D5603. Differential Revision: https://reviews.freebsd.org/D9321
2017-01-24 22:46:43 +00:00
}
Moved elf brand identification into a function. Fully identify the brand early in the process of loading an elf file, so that we can identify the sysentvec, and so that we do not continue if we do not have a brand (and thus a sysentvec). Use the values in the sysentvec for the page size and vm ranges unconditionally, since they are all filled in now.
2002-09-02 04:50:57 +00:00
Maintain the lock on the vnode for most of exec_elfN_imgact(). Specifically, it is required for the I/O that may be performed by elfN_load_section(). Avoid an obscure deadlock in the a.out, elf, and gzip image activators. Add a comment describing why the deadlock does not occur in the common case and how it might occur in less usual circumstances. Eliminate an unused variable from exec_aout_imgact(). In collaboration with: tegge
2005-12-24 04:57:50 +00:00
/*
* Avoid a possible deadlock if the current address space is destroyed
* and that address space maps the locked vnode. In the common case,
* the locked vnode's v_usecount is decremented but remains greater
* than zero. Consequently, the vnode lock is not needed by vrele().
* However, in cases where the vnode lock is external, such as nullfs,
* v_usecount may become zero.
Explain why it is safe to unlock the vnode. Requested by: kib
2012-01-17 16:20:50 +00:00
*
* The VV_TEXT flag prevents modifications to the executable while
* the vnode is unlocked.
Maintain the lock on the vnode for most of exec_elfN_imgact(). Specifically, it is required for the I/O that may be performed by elfN_load_section(). Avoid an obscure deadlock in the a.out, elf, and gzip image activators. Add a comment describing why the deadlock does not occur in the common case and how it might occur in less usual circumstances. Eliminate an unused variable from exec_aout_imgact(). In collaboration with: tegge
2005-12-24 04:57:50 +00:00
*/
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(imgp->vp, 0);
Maintain the lock on the vnode for most of exec_elfN_imgact(). Specifically, it is required for the I/O that may be performed by elfN_load_section(). Avoid an obscure deadlock in the a.out, elf, and gzip image activators. Add a comment describing why the deadlock does not occur in the common case and how it might occur in less usual circumstances. Eliminate an unused variable from exec_aout_imgact(). In collaboration with: tegge
2005-12-24 04:57:50 +00:00
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
/*
* Decide whether to enable randomization of user mappings.
* First, reset user preferences for the setid binaries.
* Then, account for the support of the randomization by the
* ABI, by user preferences, and make special treatment for
* PIE binaries.
*/
if (imgp->credential_setid) {
PROC_LOCK(imgp->proc);
imgp->proc->p_flag2 &= ~(P2_ASLR_ENABLE | P2_ASLR_DISABLE);
PROC_UNLOCK(imgp->proc);
}
if ((sv->sv_flags & SV_ASLR) == 0 ||
(imgp->proc->p_flag2 & P2_ASLR_DISABLE) != 0 ||
(fctl0 & NT_FREEBSD_FCTL_ASLR_DISABLE) != 0) {
KASSERT(et_dyn_addr != ET_DYN_ADDR_RAND,
("et_dyn_addr == RAND and !ASLR"));
} else if ((imgp->proc->p_flag2 & P2_ASLR_ENABLE) != 0 ||
(__elfN(aslr_enabled) && hdr->e_type == ET_EXEC) ||
et_dyn_addr == ET_DYN_ADDR_RAND) {
imgp->map_flags |= MAP_ASLR;
/*
* If user does not care about sbrk, utilize the bss
* grow region for mappings as well. We can select
* the base for the image anywere and still not suffer
* from the fragmentation.
*/
if (!__elfN(aslr_honor_sbrk) ||
(imgp->proc->p_flag2 & P2_ASLR_IGNSTART) != 0)
imgp->map_flags |= MAP_ASLR_IGNSTART;
}
Fix for the panic("vm_thread_new: kstack allocation failed") and silent NULL pointer dereference in the i386 and sparc64 pmap_pinit() when the kmem_alloc_nofault() failed to allocate address space. Both functions now return error instead of panicing or dereferencing NULL. As consequence, vmspace_exec() and vmspace_unshare() returns the errno int. struct vmspace arg was added to vm_forkproc() to avoid dealing with failed allocation when most of the fork1() job is already done. The kernel stack for the thread is now set up in the thread_alloc(), that itself may return NULL. Also, allocation of the first process thread is performed in the fork1() to properly deal with stack allocation failure. proc_linkup() is separated into proc_linkup() called from fork1(), and proc_linkup0(), that is used to set up the kernel process (was known as swapper). In collaboration with: Peter Holm Reviewed by: jhb
2007-11-05 11:36:16 +00:00
error = exec_new_vmspace(imgp, sv);
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
vmspace = imgp->proc->p_vmspace;
map = &vmspace->vm_map;
Rework the support for ABIs to override resource limits (used by 32-bit processes under 64-bit kernels). Previously, each 32-bit process overwrote its resource limits at exec() time. The problem with this approach is that the new limits affect all child processes of the 32-bit process, including if the child process forks and execs a 64-bit process. To fix this, don't ovewrite the resource limits during exec(). Instead, sv_fixlimits() is now replaced with a different function sv_fixlimit() which asks the ABI to sanitize a single resource limit. We then use this when querying and setting resource limits. Thus, if a 32-bit process sets a limit, then that new limit will be inherited by future children. However, if the 32-bit process doesn't change a limit, then a future 64-bit child will see the "full" 64-bit limit rather than the 32-bit limit. MFC is tentative since it will break the ABI of old linux.ko modules (no other modules are affected). MFC after: 1 week
2007-05-14 22:40:04 +00:00
imgp->proc->p_sysent = sv;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
maxv = vm_map_max(map) - lim_max(td, RLIMIT_STACK);
if (et_dyn_addr == ET_DYN_ADDR_RAND) {
KASSERT((map->flags & MAP_ASLR) != 0,
("ET_DYN_ADDR_RAND but !MAP_ASLR"));
et_dyn_addr = __CONCAT(rnd_, __elfN(base))(map,
vm_map_min(map) + mapsz + lim_max(td, RLIMIT_DATA),
/* reserve half of the address space to interpreter */
maxv / 2, 1UL << flsl(maxalign));
}
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
if (error != 0)
goto ret;
Maintain the lock on the vnode for most of exec_elfN_imgact(). Specifically, it is required for the I/O that may be performed by elfN_load_section(). Avoid an obscure deadlock in the a.out, elf, and gzip image activators. Add a comment describing why the deadlock does not occur in the common case and how it might occur in less usual circumstances. Eliminate an unused variable from exec_aout_imgact(). In collaboration with: tegge
2005-12-24 04:57:50 +00:00
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
error = __elfN(load_sections)(imgp, hdr, phdr, et_dyn_addr, NULL);
if (error != 0)
goto ret;
Change the way text_addr and data_addr are computed to use the executable status of segments instead of detecting the main text segment by which segment contains the program entry point. This affects obreak() and is required for correct operation of that function on 64-bit PowerPC systems. The previous behavior was apparently required only for the Alpha, which is no longer supported. Reviewed by: jhb Tested on: amd64, sparc64, powerpc
2010-03-25 14:21:22 +00:00
Factor out resource limit enforcement code in the ELF loader. It makes the code slightly easier to follow, and might make it easier to fix the resouce accounting to also account for the interpreter. The PROC_UNLOCK() is moved earlier - I don't see anything it should protect; the lim_max() is a wrapper around lim_rlimit(), and that, differently from lim_rlimit_proc(), doesn't require the proc lock to be held. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19689
2019-03-26 15:35:49 +00:00
error = __elfN(enforce_limits)(imgp, hdr, phdr, et_dyn_addr);
if (error != 0)
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
goto ret;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Factor out resource limit enforcement code in the ELF loader. It makes the code slightly easier to follow, and might make it easier to fix the resouce accounting to also account for the interpreter. The PROC_UNLOCK() is moved earlier - I don't see anything it should protect; the lim_max() is a wrapper around lim_rlimit(), and that, differently from lim_rlimit_proc(), doesn't require the proc lock to be held. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19689
2019-03-26 15:35:49 +00:00
entry = (u_long)hdr->e_entry + et_dyn_addr;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Add sysentvec->sv_fixlimits() hook so that we can catch cases on 64 bit systems where the data/stack/etc limits are too big for a 32 bit process. Move the 5 or so identical instances of ELF_RTLD_ADDR() into imgact_elf.c. Supply an ia32_fixlimits function. Export the clip/default values to sysctl under the compat.ia32 heirarchy. Have mmap(0, ...) respect the current p->p_limits[RLIMIT_DATA].rlim_max value rather than the sysctl tweakable variable. This allows mmap to place mappings at sensible locations when limits have been reduced. Have the imgact_elf.c ld-elf.so.1 placement algorithm use the same method as mmap(0, ...) now does. Note that we cannot remove all references to the sysctl tweakable maxdsiz etc variables because /etc/login.conf specifies a datasize of 'unlimited'. And that causes exec etc to fail since it can no longer find space to mmap things.
2003-09-25 01:10:26 +00:00
/*
* We load the dynamic linker where a userland call
* to mmap(0, ...) would put it. The rationale behind this
* calculation is that it leaves room for the heap to grow to
* its maximum allowed size.
*/
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
addr = round_page((vm_offset_t)vmspace->vm_daddr + lim_max(td,
Improve abstraction. Eliminate direct access by elf*_load_section() to an OBJT_VNODE-specific field of the vm object. The same information can be just as easily obtained from the struct vattr that is in struct image_params if the latter is passed to elf*_load_section(). Moreover, by replacing the vmspace and vm object parameters to elf*_load_section() with a struct image_params parameter, we actually reduce the size of the object code. In collaboration with: kib
2012-01-17 00:27:32 +00:00
RLIMIT_DATA));
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
if ((map->flags & MAP_ASLR) != 0) {
maxv1 = maxv / 2 + addr / 2;
MPASS(maxv1 >= addr); /* No overflow */
map->anon_loc = __CONCAT(rnd_, __elfN(base))(map, addr, maxv1,
MAXPAGESIZES > 1 ? pagesizes[1] : pagesizes[0]);
} else {
map->anon_loc = addr;
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Prepare kernel to take advantage of "branded" ELF binaries.
1996-10-16 17:51:08 +00:00
imgp->entry_addr = entry;
Add an additional field to the elf brandinfo structure to support quicker exec-time replacement of the elf interpreter on an emulation environment where an entire /compat/* tree isn't really warranted.
2003-12-23 02:42:39 +00:00
if (interp != NULL) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(imgp->vp, 0);
Implement Address Space Layout Randomization (ASLR) With this change, randomization can be enabled for all non-fixed mappings. It means that the base address for the mapping is selected with a guaranteed amount of entropy (bits). If the mapping was requested to be superpage aligned, the randomization honours the superpage attributes. Although the value of ASLR is diminshing over time as exploit authors work out simple ASLR bypass techniques, it elimintates the trivial exploitation of certain vulnerabilities, at least in theory. This implementation is relatively small and happens at the correct architectural level. Also, it is not expected to introduce regressions in existing cases when turned off (default for now), or cause any significant maintaince burden. The randomization is done on a best-effort basis - that is, the allocator falls back to a first fit strategy if fragmentation prevents entropy injection. It is trivial to implement a strong mode where failure to guarantee the requested amount of entropy results in mapping request failure, but I do not consider that to be usable. I have not fine-tuned the amount of entropy injected right now. It is only a quantitive change that will not change the implementation. The current amount is controlled by aslr_pages_rnd. To not spoil coalescing optimizations, to reduce the page table fragmentation inherent to ASLR, and to keep the transient superpage promotion for the malloced memory, locality clustering is implemented for anonymous private mappings, which are automatically grouped until fragmentation kicks in. The initial location for the anon group range is, of course, randomized. This is controlled by vm.cluster_anon, enabled by default. The default mode keeps the sbrk area unpopulated by other mappings, but this can be turned off, which gives much more breathing bits on architectures with small address space, such as i386. This is tied with the question of following an application's hint about the mmap(2) base address. Testing shows that ignoring the hint does not affect the function of common applications, but I would expect more demanding code could break. By default sbrk is preserved and mmap hints are satisfied, which can be changed by using the kern.elf{32,64}.aslr.honor_sbrk sysctl. ASLR is enabled on per-ABI basis, and currently it is only allowed on FreeBSD native i386 and amd64 (including compat 32bit) ABIs. Support for additional architectures will be added after further testing. Both per-process and per-image controls are implemented: - procctl(2) adds PROC_ASLR_CTL/PROC_ASLR_STATUS; - NT_FREEBSD_FCTL_ASLR_DISABLE feature control note bit makes it possible to force ASLR off for the given binary. (A tool to edit the feature control note is in development.) Global controls are: - kern.elf{32,64}.aslr.enable - for non-fixed mappings done by mmap(2); - kern.elf{32,64}.aslr.pie_enable - for PIE image activation mappings; - kern.elf{32,64}.aslr.honor_sbrk - allow to use sbrk area for mmap(2); - vm.cluster_anon - enables anon mapping clustering. PR: 208580 (exp runs) Exp-runs done by: antoine Reviewed by: markj (previous version) Discussed with: emaste Tested by: pho MFC after: 1 month Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D5603
2019-02-10 17:19:45 +00:00
if ((map->flags & MAP_ASLR) != 0) {
/* Assume that interpeter fits into 1/4 of AS */
maxv1 = maxv / 2 + addr / 2;
MPASS(maxv1 >= addr); /* No overflow */
addr = __CONCAT(rnd_, __elfN(base))(map, addr,
maxv1, PAGE_SIZE);
}
Refactor ELF interpreter loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19741
2019-04-08 14:31:07 +00:00
error = __elfN(load_interp)(imgp, brand_info, interp, &addr,
&imgp->entry_addr);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
Refactor ELF interpreter loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19741
2019-04-08 14:31:07 +00:00
if (error != 0)
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
goto ret;
When a statically linked binary is executed (or at least, one without an interpreter definition in its program header), set the auxiliary ELF argument AT_BASE to 0 rather than to the address that we would have mapped the interpreter at if there had been one. The ELF ABI specifications appear to be ambiguous as to the desired behavior in this situation, as they define AT_BASE as the base address of the interpreter, but do not mention what to do if there is none. On Solaris, AT_BASE will be set to the base address of the static binary if there is no interpreter, and on Linux, AT_BASE is set to 0. We go with the Linux semantics as they are of more immediate utility and allow the early runtime environment to know that the kernel has not mapped an interpreter, but because AT_PHDR points at the ELF header for the running binary, it is still possible to retrieve all required mapping information when the process starts should it be required. Either approach would be preferable to our current behavior of passing a pointer to an unmapped region of user memory as AT_BASE. MFC after: 3 weeks
2009-01-25 12:07:43 +00:00
} else
If ET_DYN binary has non-zero base address for some reason, honour it and do not relocate the binary to ET_DYN_LOAD_ADDR. This allows for the binary author to influence address map of the process. In particular, when the binary is actually an interpeter, this allows to have almost usual process address map. Communicate the relocation bias of the mapping for interpeter-less ET_DYN binary, that is interperter itself, in AT_BASE aux entry. This way, rtld is able to find its dynamic structure and relocate itself. Note that mapbase in the rtld is still wrong and requires further fixing. Reported and tested by: rwatson Discussed with: kan MFC after: 3 days
2009-10-18 12:57:48 +00:00
addr = et_dyn_addr;
Prepare kernel to take advantage of "branded" ELF binaries.
1996-10-16 17:51:08 +00:00
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
* Construct auxargs table (used by the fixup routine)
*/
imgact_elf: do not relock the text vnode if possible. We unlock the vnode around malloc(M_WAITOK), to make it possible for pagedaemon to flush vnode pages for us. Instead of doing it unconditionally, first try M_NOWAIT allocation, which typically succeed. Only on failure, unlock the vnode and retry with M_WAITOK. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:04:01 +00:00
elf_auxargs = malloc(sizeof(Elf_Auxargs), M_TEMP, M_NOWAIT);
if (elf_auxargs == NULL) {
VOP_UNLOCK(imgp->vp, 0);
elf_auxargs = malloc(sizeof(Elf_Auxargs), M_TEMP, M_WAITOK);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
imgact_elf: do not relock the text vnode if possible. We unlock the vnode around malloc(M_WAITOK), to make it possible for pagedaemon to flush vnode pages for us. Instead of doing it unconditionally, first try M_NOWAIT allocation, which typically succeed. Only on failure, unlock the vnode and retry with M_WAITOK. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:04:01 +00:00
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
elf_auxargs->execfd = -1;
Factor out section loading into a separate function. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19846
2019-04-09 15:24:38 +00:00
elf_auxargs->phdr = proghdr + et_dyn_addr;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
elf_auxargs->phent = hdr->e_phentsize;
elf_auxargs->phnum = hdr->e_phnum;
elf_auxargs->pagesz = PAGE_SIZE;
elf_auxargs->base = addr;
elf_auxargs->flags = 0;
elf_auxargs->entry = entry;
Export the eflags field from the elf header. This allows better discrimination between different subarch binaries, at least for mips and arm. Arm is implemented, mips is still tbd, so not currently exported. aarch64 does not export this because aarch64 binaries use different tags and flags than arm. Differential Revision: https://reviews.freebsd.org/D2611
2015-05-22 20:50:35 +00:00
elf_auxargs->hdr_eflags = hdr->e_flags;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
imgp->auxargs = elf_auxargs;
imgp->interpreted = 0;
Add the ELF relocation base to struct image_params. This will be required to correctly relocate the executable entry point's function descriptor on powerpc64.
2010-03-25 14:31:26 +00:00
imgp->reloc_base = addr;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
imgp->proc->p_osrel = osrel;
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
imgp->proc->p_fctl0 = fctl0;
Copy the e_machine and e_flags fields from the binary into an ELF core dump. In the kernel, cache the machine and flags fields from ELF header to use in the ELF header of a core dump. For gcore, the copy these fields over from the ELF header in the binary. This matters for platforms which encode ABI information in the flags field (such as o32 vs n32 on MIPS). Reviewed by: kib Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D9392
2017-02-07 20:34:03 +00:00
imgp->proc->p_elf_machine = hdr->e_machine;
imgp->proc->p_elf_flags = hdr->e_flags;
Implement fetching of the __FreeBSD_version from the ELF ABI-tag note. The value is read into the p_osrel member of the struct proc. p_osrel is set to 0 for the binaries without the note. MFC after: 3 days
2007-12-04 12:28:07 +00:00
imgact_elf: refactor et_dyn_addr calculation This simplifies the logic somewhat. It is extracted from the change in review in D5603. Differential Revision: https://reviews.freebsd.org/D9321
2017-01-24 22:46:43 +00:00
ret:
Factor out retrieving the interpreter path from the main ELF loader routine. Reviewed by: kib MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19715
2019-03-28 21:43:01 +00:00
if (free_interp)
free(interp, M_TEMP);
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (error);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
}
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
#define suword __CONCAT(suword, __ELF_WORD_SIZE)
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
int
__elfN(freebsd_fixup)(register_t **stack_base, struct image_params *imgp)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
{
This commit fixes various 64bit portability problems required for FreeBSD/alpha. The most significant item is to change the command argument to ioctl functions from int to u_long. This change brings us inline with various other BSD versions. Driver writers may like to use (__FreeBSD_version == 300003) to detect this change. The prototype FreeBSD/alpha machdep will follow in a couple of days time.
1998-06-07 17:13:14 +00:00
Elf_Auxargs *args = (Elf_Auxargs *)imgp->auxargs;
Avoid two suword() calls per auxarg entry. Instead, construct an auxargs array and copy it out all at once. Use an array of Elf_Auxinfo rather than pairs of Elf_Addr * to represent the array. This is the correct type where pairs of words just happend to work. To reduce the size of the diff, AUXARGS_ENTRY is altered to act on this array rather than introducing a new macro. Return errors on copyout() and suword() failures and handle them in the caller. Incidentally fixes AT_RANDOM and AT_EXECFN in 32-bit linux on amd64 which incorrectly used AUXARG_ENTRY instead of AUXARGS_ENTRY_32 (now removed due to the use of proper types). Reviewed by: kib Comments from: emaste, jhb Obtained from: CheriBSD Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D15485
2018-05-24 16:25:18 +00:00
Elf_Auxinfo *argarray, *pos;
Elf_Addr *base, *auxbase;
int error;
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
base = (Elf_Addr *)*stack_base;
Avoid two suword() calls per auxarg entry. Instead, construct an auxargs array and copy it out all at once. Use an array of Elf_Auxinfo rather than pairs of Elf_Addr * to represent the array. This is the correct type where pairs of words just happend to work. To reduce the size of the diff, AUXARGS_ENTRY is altered to act on this array rather than introducing a new macro. Return errors on copyout() and suword() failures and handle them in the caller. Incidentally fixes AT_RANDOM and AT_EXECFN in 32-bit linux on amd64 which incorrectly used AUXARG_ENTRY instead of AUXARGS_ENTRY_32 (now removed due to the use of proper types). Reviewed by: kib Comments from: emaste, jhb Obtained from: CheriBSD Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D15485
2018-05-24 16:25:18 +00:00
auxbase = base + imgp->args->argc + 1 + imgp->args->envc + 1;
argarray = pos = malloc(AT_COUNT * sizeof(*pos), M_TEMP,
M_WAITOK | M_ZERO);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Minor style(9) nit.
2008-12-17 16:25:20 +00:00
if (args->execfd != -1)
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
AUXARGS_ENTRY(pos, AT_BASE, args->base);
Export the eflags field from the elf header. This allows better discrimination between different subarch binaries, at least for mips and arm. Arm is implemented, mips is still tbd, so not currently exported. aarch64 does not export this because aarch64 binaries use different tags and flags than arm. Differential Revision: https://reviews.freebsd.org/D2611
2015-05-22 20:50:35 +00:00
AUXARGS_ENTRY(pos, AT_EHDRFLAGS, args->hdr_eflags);
Supply AT_EXECPATH auxinfo entry to the interpreter, both for native and compat32 binaries. Tested by: pho Reviewed by: kan
2009-03-17 12:53:28 +00:00
if (imgp->execpathp != 0)
AUXARGS_ENTRY(pos, AT_EXECPATH, imgp->execpathp);
Allow the kern.osrelease and kern.osreldate sysctl values to be set in a jail's creation parameters. This allows the kernel version to be reliably spoofed within the jail whether examined directly with sysctl or indirectly with the uname -r and -K options. The values can only be set at jail creation time, to eliminate the need for any locking when accessing the values via sysctl. The overridden values are inherited by nested jails (unless the config for the nested jails also overrides the values). There is no sanity or range checking, other than disallowing an empty release string or a zero release date, by design. The system administrator is trusted to set sane values. Setting values that are newer than the actual running kernel will likely cause compatibility problems. Differential Revision: https://reviews.freebsd.org/D1948 Relnotes: yes
2015-02-27 16:28:55 +00:00
AUXARGS_ENTRY(pos, AT_OSRELDATE,
imgp->proc->p_ucred->cr_prison->pr_osreldate);
Supply some useful information to the started image using ELF aux vectors. In particular, provide pagesize and pagesizes array, the canary value for SSP use, number of host CPUs and osreldate. Tested by: marius (sparc64) MFC after: 1 month
2010-08-17 08:55:45 +00:00
if (imgp->canary != 0) {
AUXARGS_ENTRY(pos, AT_CANARY, imgp->canary);
AUXARGS_ENTRY(pos, AT_CANARYLEN, imgp->canarylen);
}
AUXARGS_ENTRY(pos, AT_NCPUS, mp_ncpus);
if (imgp->pagesizes != 0) {
AUXARGS_ENTRY(pos, AT_PAGESIZES, imgp->pagesizes);
AUXARGS_ENTRY(pos, AT_PAGESIZESLEN, imgp->pagesizeslen);
}
Implement mechanism to export some kernel timekeeping data to usermode, using shared page. The structures and functions have vdso prefix, to indicate the intended location of the code in some future. The versioned per-algorithm data is exported in the format of struct vdso_timehands, which mostly repeats the content of in-kernel struct timehands. Usermode reading of the structure can be lockless. Compatibility export for 32bit processes on 64bit host is also provided. Kernel also provides usermode with indication about currently used timecounter, so that libc can fall back to syscall if configured timecounter is unknown to usermode code. The shared data updates are initiated both from the tc_windup(), where a fast task is queued to do the update, and from sysctl handlers which change timecounter. A manual override switch kern.timecounter.fast_gettime allows to turn off the mechanism. Only x86 architectures export the real algorithm data, and there, only for tsc timecounter. HPET counters page could be exported as well, but I prefer to not further glue the kernel and libc ABI there until proper vdso-based solution is developed. Minimal stubs neccessary for non-x86 architectures to still compile are provided. Discussed with: bde Reviewed by: jhb Tested by: flo MFC after: 1 month
2012-06-22 07:06:40 +00:00
if (imgp->sysent->sv_timekeep_base != 0) {
AUXARGS_ENTRY(pos, AT_TIMEKEEP,
imgp->sysent->sv_timekeep_base);
}
Use the same expression to report stack protection mode for AT_STACKEXEC as the expression used by exec_new_vmspace().
2011-01-08 18:41:19 +00:00
AUXARGS_ENTRY(pos, AT_STACKPROT, imgp->sysent->sv_shared_page_obj
!= NULL && imgp->stack_prot != 0 ? imgp->stack_prot :
imgp->sysent->sv_stackprot);
Add AT_HWCAP and AT_EHDRFLAGS on all platforms. A new 'u_long *sv_hwcap' field is added to 'struct sysentvec'. A process ABI can set this field to point to a value holding a mask of architecture-specific CPU feature flags. If an ABI does not wish to supply AT_HWCAP to processes the field can be left as NULL. The support code for AT_EHDRFLAGS was already present on all systems, just the #define was not present. This is a step towards unifying the AT_* constants across platforms. Reviewed by: kib MFC after: 1 month Differential Revision: https://reviews.freebsd.org/D12290
2017-09-14 14:26:55 +00:00
if (imgp->sysent->sv_hwcap != NULL)
AUXARGS_ENTRY(pos, AT_HWCAP, *imgp->sysent->sv_hwcap);
Add AT_HWCAP2 ELF auxiliary vector. - allocate value for new AT_HWCAP2 auxiliary vector on all platforms. - expand 'struct sysentvec' by new 'u_long *sv_hwcap2', in exactly same way as for AT_HWCAP. MFC after: 1 month Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D12699
2017-10-21 12:05:01 +00:00
if (imgp->sysent->sv_hwcap2 != NULL)
AUXARGS_ENTRY(pos, AT_HWCAP2, *imgp->sysent->sv_hwcap2);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
AUXARGS_ENTRY(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
Correct pointer subtraction in KASSERT(). The assertion would never fire without truly spectacular future programming errors. Reported by: Coverity CID: 1391367, 1391368 Sponsored by: DARPA, AFRL
2018-05-29 17:49:03 +00:00
KASSERT(pos - argarray <= AT_COUNT, ("Too many auxargs"));
Avoid two suword() calls per auxarg entry. Instead, construct an auxargs array and copy it out all at once. Use an array of Elf_Auxinfo rather than pairs of Elf_Addr * to represent the array. This is the correct type where pairs of words just happend to work. To reduce the size of the diff, AUXARGS_ENTRY is altered to act on this array rather than introducing a new macro. Return errors on copyout() and suword() failures and handle them in the caller. Incidentally fixes AT_RANDOM and AT_EXECFN in 32-bit linux on amd64 which incorrectly used AUXARG_ENTRY instead of AUXARGS_ENTRY_32 (now removed due to the use of proper types). Reviewed by: kib Comments from: emaste, jhb Obtained from: CheriBSD Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D15485
2018-05-24 16:25:18 +00:00
error = copyout(argarray, auxbase, sizeof(*argarray) * AT_COUNT);
free(argarray, M_TEMP);
if (error != 0)
return (error);
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
base--;
Avoid two suword() calls per auxarg entry. Instead, construct an auxargs array and copy it out all at once. Use an array of Elf_Auxinfo rather than pairs of Elf_Addr * to represent the array. This is the correct type where pairs of words just happend to work. To reduce the size of the diff, AUXARGS_ENTRY is altered to act on this array rather than introducing a new macro. Return errors on copyout() and suword() failures and handle them in the caller. Incidentally fixes AT_RANDOM and AT_EXECFN in 32-bit linux on amd64 which incorrectly used AUXARG_ENTRY instead of AUXARGS_ENTRY_32 (now removed due to the use of proper types). Reviewed by: kib Comments from: emaste, jhb Obtained from: CheriBSD Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D15485
2018-05-24 16:25:18 +00:00
if (suword(base, imgp->args->argc) == -1)
return (EFAULT);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
*stack_base = (register_t *)base;
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (0);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
/*
* Code for generating ELF core dumps.
*/
Remove __P.
2002-03-19 21:25:46 +00:00
typedef void (*segment_callback)(vm_map_entry_t, void *);
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
/* Closure for cb_put_phdr(). */
struct phdr_closure {
Elf_Phdr *phdr; /* Program header to fill in */
Elf_Off offset; /* Offset of segment in core file */
};
/* Closure for cb_size_segment(). */
struct sseg_closure {
int count; /* Count of writable segments. */
size_t size; /* Total size of all writable segments. */
};
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
typedef void (*outfunc_t)(void *, struct sbuf *, size_t *);
struct note_info {
int type; /* Note type. */
outfunc_t outfunc; /* Output function. */
void *outarg; /* Argument for the output function. */
size_t outsize; /* Output size. */
TAILQ_ENTRY(note_info) link; /* Link to the next note info. */
};
TAILQ_HEAD(note_info_list, note_info);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
/* Coredump output parameters. */
struct coredump_params {
off_t offset;
struct ucred *active_cred;
struct ucred *file_cred;
struct thread *td;
struct vnode *vp;
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
struct compressor *comp;
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
};
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
extern int compress_user_cores;
extern int compress_user_cores_level;
Remove __P.
2002-03-19 21:25:46 +00:00
static void cb_put_phdr(vm_map_entry_t, void *);
static void cb_size_segment(vm_map_entry_t, void *);
don't abort writing of a core dump after EFAULT It's possible to get EFAULT when writing a segment backed by a file if the segment extends beyond the file. The core dump could still be useful if we skip the rest of the segment and proceed to other segements. The skipped segment (or a portion of it) will be zero-filled. While there, use 'const' to signify that core_write() only reads the buffer and use __DECONST before calling vn_rdwr_inchunks() because it can be used for both reading and writing. Before the change: kernel: Failed to write core file for process mmap_trunc_core (error 14) kernel: pid 77718 (mmap_trunc_core), uid 1001: exited on signal 6 After the change: kernel: Failed to fully fault in a core file segment at VA 0x800645000 with size 0x4000 to be written at offset 0x29000 for process mmap_trunc_core kernel: pid 4901 (mmap_trunc_core), uid 1001: exited on signal 6 (core dumped) Reviewed by: julian, kib Obtained from: Panzura (older version of the change) MFC after: 5 days Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D9233
2017-01-20 13:39:07 +00:00
static int core_write(struct coredump_params *, const void *, size_t, off_t,
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
enum uio_seg);
imgact_elf: Rename the segment iterator to match reality The each_writable_segment routine evaluates segments on a slightly little more nuanced metric than simply "writable" or not. Rename the function to more closely match its behavior (each_dumpable_segment). Suggested by: jhb Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:51:33 +00:00
static void each_dumpable_segment(struct thread *, segment_callback, void *);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
static int __elfN(corehdr)(struct coredump_params *, int, void *, size_t,
struct note_info_list *, size_t);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
static void __elfN(prepare_notes)(struct thread *, struct note_info_list *,
size_t *);
static void __elfN(puthdr)(struct thread *, void *, size_t, int, size_t);
static void __elfN(putnote)(struct note_info *, struct sbuf *);
static size_t register_note(struct note_info_list *, int, outfunc_t, void *);
static int sbuf_drain_core_output(void *, const char *, int);
static void __elfN(note_fpregset)(void *, struct sbuf *, size_t *);
static void __elfN(note_prpsinfo)(void *, struct sbuf *, size_t *);
static void __elfN(note_prstatus)(void *, struct sbuf *, size_t *);
static void __elfN(note_threadmd)(void *, struct sbuf *, size_t *);
static void __elfN(note_thrmisc)(void *, struct sbuf *, size_t *);
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
static void __elfN(note_ptlwpinfo)(void *, struct sbuf *, size_t *);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
static void __elfN(note_procstat_auxv)(void *, struct sbuf *, size_t *);
static void __elfN(note_procstat_proc)(void *, struct sbuf *, size_t *);
static void __elfN(note_procstat_psstrings)(void *, struct sbuf *, size_t *);
static void note_procstat_files(void *, struct sbuf *, size_t *);
static void note_procstat_groups(void *, struct sbuf *, size_t *);
static void note_procstat_osrel(void *, struct sbuf *, size_t *);
static void note_procstat_rlimit(void *, struct sbuf *, size_t *);
static void note_procstat_umask(void *, struct sbuf *, size_t *);
static void note_procstat_vmmap(void *, struct sbuf *, size_t *);
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
/*
* Write out a core segment to the compression stream.
*/
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
static int
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
compress_chunk(struct coredump_params *p, char *base, char *buf, u_int len)
{
u_int chunk_len;
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
int error;
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
while (len > 0) {
chunk_len = MIN(len, CORE_BUF_SIZE);
don't abort writing of a core dump after EFAULT It's possible to get EFAULT when writing a segment backed by a file if the segment extends beyond the file. The core dump could still be useful if we skip the rest of the segment and proceed to other segements. The skipped segment (or a portion of it) will be zero-filled. While there, use 'const' to signify that core_write() only reads the buffer and use __DECONST before calling vn_rdwr_inchunks() because it can be used for both reading and writing. Before the change: kernel: Failed to write core file for process mmap_trunc_core (error 14) kernel: pid 77718 (mmap_trunc_core), uid 1001: exited on signal 6 After the change: kernel: Failed to fully fault in a core file segment at VA 0x800645000 with size 0x4000 to be written at offset 0x29000 for process mmap_trunc_core kernel: pid 4901 (mmap_trunc_core), uid 1001: exited on signal 6 (core dumped) Reviewed by: julian, kib Obtained from: Panzura (older version of the change) MFC after: 5 days Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D9233
2017-01-20 13:39:07 +00:00
/*
* We can get EFAULT error here.
* In that case zero out the current chunk of the segment.
*/
error = copyin(base, buf, chunk_len);
if (error != 0)
bzero(buf, chunk_len);
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
error = compressor_write(p->comp, buf, chunk_len);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
if (error != 0)
break;
base += chunk_len;
len -= chunk_len;
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
}
return (error);
}
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
static int
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
core_compressed_write(void *base, size_t len, off_t offset, void *arg)
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
{
return (core_write((struct coredump_params *)arg, base, len, offset,
UIO_SYSSPACE));
}
static int
don't abort writing of a core dump after EFAULT It's possible to get EFAULT when writing a segment backed by a file if the segment extends beyond the file. The core dump could still be useful if we skip the rest of the segment and proceed to other segements. The skipped segment (or a portion of it) will be zero-filled. While there, use 'const' to signify that core_write() only reads the buffer and use __DECONST before calling vn_rdwr_inchunks() because it can be used for both reading and writing. Before the change: kernel: Failed to write core file for process mmap_trunc_core (error 14) kernel: pid 77718 (mmap_trunc_core), uid 1001: exited on signal 6 After the change: kernel: Failed to fully fault in a core file segment at VA 0x800645000 with size 0x4000 to be written at offset 0x29000 for process mmap_trunc_core kernel: pid 4901 (mmap_trunc_core), uid 1001: exited on signal 6 (core dumped) Reviewed by: julian, kib Obtained from: Panzura (older version of the change) MFC after: 5 days Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D9233
2017-01-20 13:39:07 +00:00
core_write(struct coredump_params *p, const void *base, size_t len,
off_t offset, enum uio_seg seg)
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
{
don't abort writing of a core dump after EFAULT It's possible to get EFAULT when writing a segment backed by a file if the segment extends beyond the file. The core dump could still be useful if we skip the rest of the segment and proceed to other segements. The skipped segment (or a portion of it) will be zero-filled. While there, use 'const' to signify that core_write() only reads the buffer and use __DECONST before calling vn_rdwr_inchunks() because it can be used for both reading and writing. Before the change: kernel: Failed to write core file for process mmap_trunc_core (error 14) kernel: pid 77718 (mmap_trunc_core), uid 1001: exited on signal 6 After the change: kernel: Failed to fully fault in a core file segment at VA 0x800645000 with size 0x4000 to be written at offset 0x29000 for process mmap_trunc_core kernel: pid 4901 (mmap_trunc_core), uid 1001: exited on signal 6 (core dumped) Reviewed by: julian, kib Obtained from: Panzura (older version of the change) MFC after: 5 days Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D9233
2017-01-20 13:39:07 +00:00
return (vn_rdwr_inchunks(UIO_WRITE, p->vp, __DECONST(void *, base),
len, offset, seg, IO_UNIT | IO_DIRECT | IO_RANGELOCKED,
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
p->active_cred, p->file_cred, NULL, p->td));
}
static int
core_output(void *base, size_t len, off_t offset, struct coredump_params *p,
void *tmpbuf)
{
don't abort writing of a core dump after EFAULT It's possible to get EFAULT when writing a segment backed by a file if the segment extends beyond the file. The core dump could still be useful if we skip the rest of the segment and proceed to other segements. The skipped segment (or a portion of it) will be zero-filled. While there, use 'const' to signify that core_write() only reads the buffer and use __DECONST before calling vn_rdwr_inchunks() because it can be used for both reading and writing. Before the change: kernel: Failed to write core file for process mmap_trunc_core (error 14) kernel: pid 77718 (mmap_trunc_core), uid 1001: exited on signal 6 After the change: kernel: Failed to fully fault in a core file segment at VA 0x800645000 with size 0x4000 to be written at offset 0x29000 for process mmap_trunc_core kernel: pid 4901 (mmap_trunc_core), uid 1001: exited on signal 6 (core dumped) Reviewed by: julian, kib Obtained from: Panzura (older version of the change) MFC after: 5 days Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D9233
2017-01-20 13:39:07 +00:00
int error;
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
if (p->comp != NULL)
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
return (compress_chunk(p, base, tmpbuf, len));
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
don't abort writing of a core dump after EFAULT It's possible to get EFAULT when writing a segment backed by a file if the segment extends beyond the file. The core dump could still be useful if we skip the rest of the segment and proceed to other segements. The skipped segment (or a portion of it) will be zero-filled. While there, use 'const' to signify that core_write() only reads the buffer and use __DECONST before calling vn_rdwr_inchunks() because it can be used for both reading and writing. Before the change: kernel: Failed to write core file for process mmap_trunc_core (error 14) kernel: pid 77718 (mmap_trunc_core), uid 1001: exited on signal 6 After the change: kernel: Failed to fully fault in a core file segment at VA 0x800645000 with size 0x4000 to be written at offset 0x29000 for process mmap_trunc_core kernel: pid 4901 (mmap_trunc_core), uid 1001: exited on signal 6 (core dumped) Reviewed by: julian, kib Obtained from: Panzura (older version of the change) MFC after: 5 days Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D9233
2017-01-20 13:39:07 +00:00
/*
* EFAULT is a non-fatal error that we can get, for example,
* if the segment is backed by a file but extends beyond its
* end.
*/
error = core_write(p, base, len, offset, UIO_USERSPACE);
if (error == EFAULT) {
log(LOG_WARNING, "Failed to fully fault in a core file segment "
"at VA %p with size 0x%zx to be written at offset 0x%jx "
"for process %s\n", base, len, offset, curproc->p_comm);
/*
* Write a "real" zero byte at the end of the target region
* in the case this is the last segment.
* The intermediate space will be implicitly zero-filled.
*/
error = core_write(p, zero_region, 1, offset + len - 1,
UIO_SYSSPACE);
}
return (error);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
/*
* Drain into a core file.
*/
static int
sbuf_drain_core_output(void *arg, const char *data, int len)
{
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
struct coredump_params *p;
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
int error, locked;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
p = (struct coredump_params *)arg;
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
/*
* Some kern_proc out routines that print to this sbuf may
* call us with the process lock held. Draining with the
* non-sleepable lock held is unsafe. The lock is needed for
* those routines when dumping a live process. In our case we
* can safely release the lock before draining and acquire
* again after.
*/
locked = PROC_LOCKED(p->td->td_proc);
if (locked)
PROC_UNLOCK(p->td->td_proc);
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
if (p->comp != NULL)
error = compressor_write(p->comp, __DECONST(char *, data), len);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
else
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
error = core_write(p, __DECONST(void *, data), len, p->offset,
UIO_SYSSPACE);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
if (locked)
PROC_LOCK(p->td->td_proc);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
if (error != 0)
return (-error);
p->offset += len;
return (len);
}
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
int
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
__elfN(coredump)(struct thread *td, struct vnode *vp, off_t limit, int flags)
General cleanup of core-dumping code. Submitted by: Sean Fagan,
1999-09-01 00:29:56 +00:00
{
Allocate TIDs in thread_init() and deallocate them in thread_fini(). The overhead of unconditionally allocating TIDs (and likewise, unconditionally deallocating them), is amortized across multiple thread creations by the way UMA makes it possible to have type-stable storage. Previously the cost was kept down by having threads created as part of a fork operation use the process' PID as the TID. While this had some nice properties, it also introduced complexity in the way TIDs were allocated. Most importantly, by using the type-stable storage that UMA gives us this was also unnecessary. This change affects how core dumps are created and in particular how the PRSTATUS notes are dumped. Since we don't have a thread with a TID equalling the PID, we now need a different way to preserve the old and previous behavior. We do this by having the given thread (i.e. the thread passed to the core dump code in td) dump it's state first and fill in pr_pid with the actual PID. All other threads will have pr_pid contain their TIDs. The upshot of all this is that the debugger will now likely select the right LWP (=TID) as the initial thread. Credits to: julian@ for spotting how we can utilize UMA. Thanks to: all who provided julian@ with test results.
2004-06-26 18:58:22 +00:00
struct ucred *cred = td->td_ucred;
General cleanup of core-dumping code. Submitted by: Sean Fagan,
1999-09-01 00:29:56 +00:00
int error = 0;
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
struct sseg_closure seginfo;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
struct note_info_list notelst;
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
struct coredump_params params;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
struct note_info *ninfo;
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
void *hdr, *tmpbuf;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
size_t hdrsize, notesz, coresize;
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
hdr = NULL;
Fix some error-handling bugs when core dump compression is enabled: - Ensure that core dump parameters are initialized in the error path. - Don't call gzio_fini() on a NULL stream. Reported by: rpaulo
2015-07-14 18:24:05 +00:00
tmpbuf = NULL;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
TAILQ_INIT(&notelst);
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
/* Size the program segments. */
seginfo.count = 0;
seginfo.size = 0;
imgact_elf: Rename the segment iterator to match reality The each_writable_segment routine evaluates segments on a slightly little more nuanced metric than simply "writable" or not. Rename the function to more closely match its behavior (each_dumpable_segment). Suggested by: jhb Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:51:33 +00:00
each_dumpable_segment(td, cb_size_segment, &seginfo);
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
/*
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
* Collect info about the core file header area.
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
*/
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
hdrsize = sizeof(Elf_Ehdr) + sizeof(Elf_Phdr) * (1 + seginfo.count);
Extend ELF coredump to support more than 65535 segments The ELF e_phnum field is only 16 bits wide. To support more than 65535 segments (program headers), Sun's "Linker and Libraries Guide" table 7-7 (or 12-7, depending on document version) prescribes a special first section header where sh_info represents the real number of program headers. Test code to follow, when it is ready. Reference: http://docs.oracle.com/cd/E18752_01/pdf/817-1984.pdf Reviewed by: emaste, markj Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D7255
2016-07-20 16:59:36 +00:00
if (seginfo.count + 1 >= PN_XNUM)
hdrsize += sizeof(Elf_Shdr);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
__elfN(prepare_notes)(td, &notelst, &notesz);
coresize = round_page(hdrsize + notesz) + seginfo.size;
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
Fix some error-handling bugs when core dump compression is enabled: - Ensure that core dump parameters are initialized in the error path. - Don't call gzio_fini() on a NULL stream. Reported by: rpaulo
2015-07-14 18:24:05 +00:00
/* Set up core dump parameters. */
params.offset = 0;
params.active_cred = cred;
params.file_cred = NOCRED;
params.td = td;
params.vp = vp;
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
params.comp = NULL;
Fix some error-handling bugs when core dump compression is enabled: - Ensure that core dump parameters are initialized in the error path. - Don't call gzio_fini() on a NULL stream. Reported by: rpaulo
2015-07-14 18:24:05 +00:00
All the racct_*() calls need to happen with the proc locked. Fixing this won't happen before 9.0. This commit adds "#ifdef RACCT" around all the "PROC_LOCK(p); racct_whatever(p, ...); PROC_UNLOCK(p)" instances, in order to avoid useless locking/unlocking in kernels built without "options RACCT".
2011-07-06 20:06:44 +00:00
#ifdef RACCT
Add kern.racct.enable tunable and RACCT_DISABLED config option. The point of this is to be able to add RACCT (with RACCT_DISABLED) to GENERIC, to avoid having to rebuild the kernel to use rctl(8). Differential Revision: https://reviews.freebsd.org/D2369 Reviewed by: kib@ MFC after: 1 month Relnotes: yes Sponsored by: The FreeBSD Foundation
2015-04-29 10:23:02 +00:00
if (racct_enable) {
PROC_LOCK(td->td_proc);
error = racct_add(td->td_proc, RACCT_CORE, coresize);
PROC_UNLOCK(td->td_proc);
if (error != 0) {
error = EFAULT;
goto done;
}
Add accounting for most of the memory-related resources. Sponsored by: The FreeBSD Foundation Reviewed by: kib (earlier version)
2011-04-05 20:23:59 +00:00
}
All the racct_*() calls need to happen with the proc locked. Fixing this won't happen before 9.0. This commit adds "#ifdef RACCT" around all the "PROC_LOCK(p); racct_whatever(p, ...); PROC_UNLOCK(p)" instances, in order to avoid useless locking/unlocking in kernels built without "options RACCT".
2011-07-06 20:06:44 +00:00
#endif
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
if (coresize >= limit) {
Don't leak core_buf or gzfile if doing a compressed core file and we hit an error condition. Obtained from: Juniper Networks
2010-04-30 03:13:24 +00:00
error = EFAULT;
goto done;
}
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
/* Create a compression stream if necessary. */
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
if (compress_user_cores != 0) {
params.comp = compressor_init(core_compressed_write,
compress_user_cores, CORE_BUF_SIZE,
compress_user_cores_level, &params);
if (params.comp == NULL) {
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
error = EFAULT;
goto done;
}
tmpbuf = malloc(CORE_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO);
}
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
/*
* Allocate memory for building the header, fill it up,
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
* and write it out following the notes.
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
*/
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
hdr = malloc(hdrsize, M_TEMP, M_WAITOK);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
error = __elfN(corehdr)(&params, seginfo.count, hdr, hdrsize, &notelst,
notesz);
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
/* Write the contents of all of the writable segments. */
if (error == 0) {
Elf_Phdr *php;
Back out workaround for vn_rdwr_inchunks()'s INT_MAX length limitation after discussions with bde; vn_rdwr_inchunks() itself should be fixed.
2004-06-05 02:00:12 +00:00
off_t offset;
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
int i;
php = (Elf_Phdr *)((char *)hdr + sizeof(Elf_Ehdr)) + 1;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
offset = round_page(hdrsize + notesz);
Fixed white space around operators, casts and reserved words. Reviewed by: md5
2002-08-24 22:55:16 +00:00
for (i = 0; i < seginfo.count; i++) {
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
error = core_output((caddr_t)(uintptr_t)php->p_vaddr,
php->p_filesz, offset, &params, tmpbuf);
Back out workaround for vn_rdwr_inchunks()'s INT_MAX length limitation after discussions with bde; vn_rdwr_inchunks() itself should be fixed.
2004-06-05 02:00:12 +00:00
if (error != 0)
break;
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
offset += php->p_filesz;
php++;
}
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
if (error == 0 && params.comp != NULL)
error = compressor_flush(params.comp);
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
}
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
if (error) {
log(LOG_WARNING,
"Failed to write core file for process %s (error %d)\n",
curproc->p_comm, error);
}
done:
Generalize the gzio API. We currently use a set of subroutines in kern_gzio.c to perform compression of user and kernel core dumps. In the interest of adding support for other compression algorithms (zstd) in this role without complicating the API consumers, add a simple compressor API which can be used to select an algorithm. Also change the (non-default) GZIO kernel option to not enable compressed user cores by default. It's not clear that such a default would be desirable with support for multiple algorithms implemented, and it's inconsistent in that it isn't applied to kernel dumps. Reviewed by: cem Differential Revision: https://reviews.freebsd.org/D13632
2018-01-08 21:27:41 +00:00
free(tmpbuf, M_TEMP);
if (params.comp != NULL)
compressor_fini(params.comp);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
while ((ninfo = TAILQ_FIRST(&notelst)) != NULL) {
TAILQ_REMOVE(&notelst, ninfo, link);
free(ninfo, M_TEMP);
}
if (hdr != NULL)
free(hdr, M_TEMP);
Infrastructure tweaks to allow having both an Elf32 and an Elf64 executable handler in the kernel at the same time. Also, allow for the exec_new_vmspace() code to build a different sized vmspace depending on the executable environment. This is a big help for execing i386 binaries on ia64. The ELF exec code grows the ability to map partial pages when there is a page size difference, eg: emulating 4K pages on 8K or 16K hardware pages. Flesh out the i386 emulation support for ia64. At this point, the only binary that I know of that fails is cvsup, because the cvsup runtime tries to execute code in pages not marked executable. Obtained from: dfr (mostly, many tweaks from me).
2002-07-20 02:56:12 +00:00
return x; -> return (x); return(x); -> return (x); Reviewed by: md5
2002-08-24 22:01:40 +00:00
return (error);
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
}
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
/*
imgact_elf: Rename the segment iterator to match reality The each_writable_segment routine evaluates segments on a slightly little more nuanced metric than simply "writable" or not. Rename the function to more closely match its behavior (each_dumpable_segment). Suggested by: jhb Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:51:33 +00:00
* A callback for each_dumpable_segment() to write out the segment's
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
* program header entry.
*/
static void
ANSIfy sys/kern/imgact_*
2018-03-12 15:45:50 +00:00
cb_put_phdr(vm_map_entry_t entry, void *closure)
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
{
struct phdr_closure *phc = (struct phdr_closure *)closure;
Elf_Phdr *phdr = phc->phdr;
phc->offset = round_page(phc->offset);
phdr->p_type = PT_LOAD;
phdr->p_offset = phc->offset;
phdr->p_vaddr = entry->start;
phdr->p_paddr = 0;
phdr->p_filesz = phdr->p_memsz = entry->end - entry->start;
phdr->p_align = PAGE_SIZE;
Collect code to translate between vm_prot_t and p_flags into helper functions. MFC after: 1 week
2011-01-08 16:02:14 +00:00
phdr->p_flags = __elfN(untrans_prot)(entry->protection);
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
phc->offset += phdr->p_filesz;
phc->phdr++;
}
/*
imgact_elf: Rename the segment iterator to match reality The each_writable_segment routine evaluates segments on a slightly little more nuanced metric than simply "writable" or not. Rename the function to more closely match its behavior (each_dumpable_segment). Suggested by: jhb Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:51:33 +00:00
* A callback for each_dumpable_segment() to gather information about
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
* the number of segments and their total size.
*/
static void
ANSI-fy imgact_elf.c Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:46:56 +00:00
cb_size_segment(vm_map_entry_t entry, void *closure)
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
{
struct sseg_closure *ssc = (struct sseg_closure *)closure;
ssc->count++;
ssc->size += entry->end - entry->start;
}
/*
* For each writable segment in the process's memory map, call the given
* function with a pointer to the map entry and some arbitrary
* caller-supplied data.
*/
static void
imgact_elf: Rename the segment iterator to match reality The each_writable_segment routine evaluates segments on a slightly little more nuanced metric than simply "writable" or not. Rename the function to more closely match its behavior (each_dumpable_segment). Suggested by: jhb Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:51:33 +00:00
each_dumpable_segment(struct thread *td, segment_callback func, void *closure)
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
{
Allocate TIDs in thread_init() and deallocate them in thread_fini(). The overhead of unconditionally allocating TIDs (and likewise, unconditionally deallocating them), is amortized across multiple thread creations by the way UMA makes it possible to have type-stable storage. Previously the cost was kept down by having threads created as part of a fork operation use the process' PID as the TID. While this had some nice properties, it also introduced complexity in the way TIDs were allocated. Most importantly, by using the type-stable storage that UMA gives us this was also unnecessary. This change affects how core dumps are created and in particular how the PRSTATUS notes are dumped. Since we don't have a thread with a TID equalling the PID, we now need a different way to preserve the old and previous behavior. We do this by having the given thread (i.e. the thread passed to the core dump code in td) dump it's state first and fill in pr_pid with the actual PID. All other threads will have pr_pid contain their TIDs. The upshot of all this is that the debugger will now likely select the right LWP (=TID) as the initial thread. Credits to: julian@ for spotting how we can utilize UMA. Thanks to: all who provided julian@ with test results.
2004-06-26 18:58:22 +00:00
struct proc *p = td->td_proc;
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
vm_map_t map = &p->p_vmspace->vm_map;
vm_map_entry_t entry;
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
vm_object_t backing_object, object;
boolean_t ignore_entry;
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
vm_map_lock_read(map);
Define macro VM_MAP_ENTRY_FOREACH for enumerating the entries in a vm_map. In case the implementation ever changes from using a chain of next pointers, then changing the macro definition will be necessary, but changing all the files that iterate over vm_map entries will not. Drop a counter in vm_object.c that would have an effect only if the vm_map entry count was wrong. Discussed with: alc Reviewed by: markj Tested by: pho (earlier version) Differential Revision: https://reviews.freebsd.org/D21882
2019-10-08 07:14:21 +00:00
VM_MAP_ENTRY_FOREACH(entry, map) {
Change the way ELF coredumps are handled. Instead of unconditionally skipping read-only pages, which can result in valuable non-text-related data not getting dumped, the ELF loader and the dynamic loader now mark read-only text pages NOCORE and the coredump code only checks (primarily) for complete inaccessibility of the page or NOCORE being set. Certain applications which map large amounts of read-only data will produce much larger cores. A new sysctl has been added, debug.elf_legacy_coredump, which will revert to the old behavior. This commit represents collaborative work by all parties involved. The PR contains a program demonstrating the problem. PR: kern/45994 Submitted by: "Peter Edwards" <pmedwards@eircom.net>, Archie Cobbs <archie@dellroad.org> Reviewed by: jdp, dillon MFC after: 7 days
2002-12-16 19:24:43 +00:00
/*
* Don't dump inaccessible mappings, deal with legacy
* coredump mode.
*
* Note that read-only segments related to the elf binary
* are marked MAP_ENTRY_NOCOREDUMP now so we no longer
* need to arbitrarily ignore such segments.
*/
if (elf_legacy_coredump) {
if ((entry->protection & VM_PROT_RW) != VM_PROT_RW)
continue;
} else {
if ((entry->protection & VM_PROT_ALL) == 0)
continue;
}
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
Add MAP_NOCORE to mmap(2), and MADV_NOCORE and MADV_CORE to madvise(2). This This feature allows you to specify if mmap'd data is included in an application's corefile. Change the type of eflags in struct vm_map_entry from u_char to vm_eflags_t (an unsigned int). Reviewed by: dillon,jdp,alfred Approved by: jkh
2000-02-28 04:10:35 +00:00
/*
Change the way ELF coredumps are handled. Instead of unconditionally skipping read-only pages, which can result in valuable non-text-related data not getting dumped, the ELF loader and the dynamic loader now mark read-only text pages NOCORE and the coredump code only checks (primarily) for complete inaccessibility of the page or NOCORE being set. Certain applications which map large amounts of read-only data will produce much larger cores. A new sysctl has been added, debug.elf_legacy_coredump, which will revert to the old behavior. This commit represents collaborative work by all parties involved. The PR contains a program demonstrating the problem. PR: kern/45994 Submitted by: "Peter Edwards" <pmedwards@eircom.net>, Archie Cobbs <archie@dellroad.org> Reviewed by: jdp, dillon MFC after: 7 days
2002-12-16 19:24:43 +00:00
* Dont include memory segment in the coredump if
* MAP_NOCORE is set in mmap(2) or MADV_NOCORE in
* madvise(2). Do not dump submaps (i.e. parts of the
* kernel map).
*/
if (entry->eflags & (MAP_ENTRY_NOCOREDUMP|MAP_ENTRY_IS_SUB_MAP))
Add MAP_NOCORE to mmap(2), and MADV_NOCORE and MADV_CORE to madvise(2). This This feature allows you to specify if mmap'd data is included in an application's corefile. Change the type of eflags in struct vm_map_entry from u_char to vm_eflags_t (an unsigned int). Reviewed by: dillon,jdp,alfred Approved by: jkh
2000-02-28 04:10:35 +00:00
continue;
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
if ((object = entry->object.vm_object) == NULL)
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
continue;
/* Ignore memory-mapped devices and such things. */
Switch some "low-hanging fruit" to acquire read lock on vmobjects rather than write locks. Sponsored by: EMC / Isilon storage division Reviewed by: alc Tested by: pho
2013-04-08 19:58:32 +00:00
VM_OBJECT_RLOCK(object);
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
while ((backing_object = object->backing_object) != NULL) {
Switch some "low-hanging fruit" to acquire read lock on vmobjects rather than write locks. Sponsored by: EMC / Isilon storage division Reviewed by: alc Tested by: pho
2013-04-08 19:58:32 +00:00
VM_OBJECT_RLOCK(backing_object);
VM_OBJECT_RUNLOCK(object);
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
object = backing_object;
}
ignore_entry = object->type != OBJT_DEFAULT &&
Include OBJT_PHYS VM objects in ELF core dumps. In particular this includes the shared page allowing debuggers to use the signal trampoline code to identify signal frames in core dumps. Differential Revision: https://reviews.freebsd.org/D1828 Reviewed by: alc, kib MFC after: 1 week
2015-02-14 17:12:31 +00:00
object->type != OBJT_SWAP && object->type != OBJT_VNODE &&
object->type != OBJT_PHYS;
Switch some "low-hanging fruit" to acquire read lock on vmobjects rather than write locks. Sponsored by: EMC / Isilon storage division Reviewed by: alc Tested by: pho
2013-04-08 19:58:32 +00:00
VM_OBJECT_RUNLOCK(object);
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
if (ignore_entry)
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
continue;
(*func)(entry, closure);
}
Add vm map and object locking to each_writable_segment(). Noticed by: jhb@ MFC after: 3 weeks
2006-11-19 23:38:59 +00:00
vm_map_unlock_read(map);
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
}
/*
* Write the core file header to the file, including padding up to
* the page boundary.
*/
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
static int
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
__elfN(corehdr)(struct coredump_params *p, int numsegs, void *hdr,
size_t hdrsize, struct note_info_list *notelst, size_t notesz)
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
{
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
struct note_info *ninfo;
struct sbuf *sb;
int error;
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
/* Fill in the header. */
Restore the core-dumping of all writable segments for ELF executables, minus the NULL pointer dereference in rev. 1.33. Also simplify things somewhat by eliminating one traversal of the VM map entries. Finally, eliminate calls to vm_map_{un,}lock_read() which aren't needed here. I originally took them from procfs_map.c, but here we know we are dealing only with the map of the current process.
1998-09-16 02:04:05 +00:00
bzero(hdr, hdrsize);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
__elfN(puthdr)(p->td, hdr, hdrsize, numsegs, notesz);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
sb = sbuf_new(NULL, NULL, CORE_BUF_SIZE, SBUF_FIXEDLEN);
Reimplement support for userland core dump compression using a new interface in kern_gzio.c. The old gzio interface was somewhat inflexible and has not worked properly since r272535: currently, the gzio functions are called with a range lock held on the output vnode, but kern_gzio.c does not pass the IO_RANGELOCKED flag to vn_rdwr() calls, resulting in deadlock when vn_rdwr() attempts to reacquire the range lock. Moreover, the new gzio interface can be used to implement kernel core compression. This change also modifies the kernel configuration options needed to enable userland core dump compression support: gzio is now an option rather than a device, and the COMPRESS_USER_CORES option is removed. Core dump compression is enabled using the kern.compress_user_cores sysctl/tunable. Differential Revision: https://reviews.freebsd.org/D1832 Reviewed by: rpaulo Discussed with: kib
2015-03-09 03:50:53 +00:00
sbuf_set_drain(sb, sbuf_drain_core_output, p);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
sbuf_start_section(sb, NULL);
sbuf_bcat(sb, hdr, hdrsize);
TAILQ_FOREACH(ninfo, notelst, link)
__elfN(putnote)(ninfo, sb);
/* Align up to a page boundary for the program segments. */
sbuf_end_section(sb, -1, PAGE_SIZE, 0);
error = sbuf_finish(sb);
sbuf_delete(sb);
return (error);
}
static void
__elfN(prepare_notes)(struct thread *td, struct note_info_list *list,
size_t *sizep)
{
struct proc *p;
struct thread *thr;
size_t size;
p = td->td_proc;
size = 0;
size += register_note(list, NT_PRPSINFO, __elfN(note_prpsinfo), p);
/*
* To have the debugger select the right thread (LWP) as the initial
* thread, we dump the state of the thread passed to us in td first.
* This is the thread that causes the core dump and thus likely to
* be the right thread one wants to have selected in the debugger.
*/
thr = td;
while (thr != NULL) {
size += register_note(list, NT_PRSTATUS,
__elfN(note_prstatus), thr);
size += register_note(list, NT_FPREGSET,
__elfN(note_fpregset), thr);
size += register_note(list, NT_THRMISC,
__elfN(note_thrmisc), thr);
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
size += register_note(list, NT_PTLWPINFO,
__elfN(note_ptlwpinfo), thr);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
size += register_note(list, -1,
__elfN(note_threadmd), thr);
thr = (thr == td) ? TAILQ_FIRST(&p->p_threads) :
TAILQ_NEXT(thr, td_plist);
if (thr == td)
thr = TAILQ_NEXT(thr, td_plist);
}
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
size += register_note(list, NT_PROCSTAT_PROC,
__elfN(note_procstat_proc), p);
size += register_note(list, NT_PROCSTAT_FILES,
note_procstat_files, p);
size += register_note(list, NT_PROCSTAT_VMMAP,
note_procstat_vmmap, p);
size += register_note(list, NT_PROCSTAT_GROUPS,
note_procstat_groups, p);
size += register_note(list, NT_PROCSTAT_UMASK,
note_procstat_umask, p);
size += register_note(list, NT_PROCSTAT_RLIMIT,
note_procstat_rlimit, p);
size += register_note(list, NT_PROCSTAT_OSREL,
note_procstat_osrel, p);
size += register_note(list, NT_PROCSTAT_PSSTRINGS,
__elfN(note_procstat_psstrings), p);
size += register_note(list, NT_PROCSTAT_AUXV,
__elfN(note_procstat_auxv), p);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
*sizep = size;
}
static void
__elfN(puthdr)(struct thread *td, void *hdr, size_t hdrsize, int numsegs,
size_t notesz)
{
Elf_Ehdr *ehdr;
Elf_Phdr *phdr;
Extend ELF coredump to support more than 65535 segments The ELF e_phnum field is only 16 bits wide. To support more than 65535 segments (program headers), Sun's "Linker and Libraries Guide" table 7-7 (or 12-7, depending on document version) prescribes a special first section header where sh_info represents the real number of program headers. Test code to follow, when it is ready. Reference: http://docs.oracle.com/cd/E18752_01/pdf/817-1984.pdf Reviewed by: emaste, markj Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D7255
2016-07-20 16:59:36 +00:00
Elf_Shdr *shdr;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
struct phdr_closure phc;
ehdr = (Elf_Ehdr *)hdr;
ehdr->e_ident[EI_MAG0] = ELFMAG0;
ehdr->e_ident[EI_MAG1] = ELFMAG1;
ehdr->e_ident[EI_MAG2] = ELFMAG2;
ehdr->e_ident[EI_MAG3] = ELFMAG3;
ehdr->e_ident[EI_CLASS] = ELF_CLASS;
ehdr->e_ident[EI_DATA] = ELF_DATA;
ehdr->e_ident[EI_VERSION] = EV_CURRENT;
ehdr->e_ident[EI_OSABI] = ELFOSABI_FREEBSD;
ehdr->e_ident[EI_ABIVERSION] = 0;
ehdr->e_ident[EI_PAD] = 0;
ehdr->e_type = ET_CORE;
Copy the e_machine and e_flags fields from the binary into an ELF core dump. In the kernel, cache the machine and flags fields from ELF header to use in the ELF header of a core dump. For gcore, the copy these fields over from the ELF header in the binary. This matters for platforms which encode ABI information in the flags field (such as o32 vs n32 on MIPS). Reviewed by: kib Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D9392
2017-02-07 20:34:03 +00:00
ehdr->e_machine = td->td_proc->p_elf_machine;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
ehdr->e_version = EV_CURRENT;
ehdr->e_entry = 0;
ehdr->e_phoff = sizeof(Elf_Ehdr);
Copy the e_machine and e_flags fields from the binary into an ELF core dump. In the kernel, cache the machine and flags fields from ELF header to use in the ELF header of a core dump. For gcore, the copy these fields over from the ELF header in the binary. This matters for platforms which encode ABI information in the flags field (such as o32 vs n32 on MIPS). Reviewed by: kib Sponsored by: DARPA / AFRL Differential Revision: https://reviews.freebsd.org/D9392
2017-02-07 20:34:03 +00:00
ehdr->e_flags = td->td_proc->p_elf_flags;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
ehdr->e_ehsize = sizeof(Elf_Ehdr);
ehdr->e_phentsize = sizeof(Elf_Phdr);
ehdr->e_shentsize = sizeof(Elf_Shdr);
ehdr->e_shstrndx = SHN_UNDEF;
Extend ELF coredump to support more than 65535 segments The ELF e_phnum field is only 16 bits wide. To support more than 65535 segments (program headers), Sun's "Linker and Libraries Guide" table 7-7 (or 12-7, depending on document version) prescribes a special first section header where sh_info represents the real number of program headers. Test code to follow, when it is ready. Reference: http://docs.oracle.com/cd/E18752_01/pdf/817-1984.pdf Reviewed by: emaste, markj Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D7255
2016-07-20 16:59:36 +00:00
if (numsegs + 1 < PN_XNUM) {
ehdr->e_phnum = numsegs + 1;
ehdr->e_shnum = 0;
} else {
ehdr->e_phnum = PN_XNUM;
ehdr->e_shnum = 1;
ehdr->e_shoff = ehdr->e_phoff +
(numsegs + 1) * ehdr->e_phentsize;
KASSERT(ehdr->e_shoff == hdrsize - sizeof(Elf_Shdr),
("e_shoff: %zu, hdrsize - shdr: %zu",
Fix DEBUG build on 64-bit arch after r303099 Reported by: Larry Rosenman <ler at lerctr.org>
2016-07-20 18:11:22 +00:00
(size_t)ehdr->e_shoff, hdrsize - sizeof(Elf_Shdr)));
Extend ELF coredump to support more than 65535 segments The ELF e_phnum field is only 16 bits wide. To support more than 65535 segments (program headers), Sun's "Linker and Libraries Guide" table 7-7 (or 12-7, depending on document version) prescribes a special first section header where sh_info represents the real number of program headers. Test code to follow, when it is ready. Reference: http://docs.oracle.com/cd/E18752_01/pdf/817-1984.pdf Reviewed by: emaste, markj Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D7255
2016-07-20 16:59:36 +00:00
shdr = (Elf_Shdr *)((char *)hdr + ehdr->e_shoff);
memset(shdr, 0, sizeof(*shdr));
/*
* A special first section is used to hold large segment and
* section counts. This was proposed by Sun Microsystems in
* Solaris and has been adopted by Linux; the standard ELF
* tools are already familiar with the technique.
*
* See table 7-7 of the Solaris "Linker and Libraries Guide"
* (or 12-7 depending on the version of the document) for more
* details.
*/
shdr->sh_type = SHT_NULL;
shdr->sh_size = ehdr->e_shnum;
shdr->sh_link = ehdr->e_shstrndx;
shdr->sh_info = numsegs + 1;
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
/*
* Fill in the program header entries.
*/
Extend ELF coredump to support more than 65535 segments The ELF e_phnum field is only 16 bits wide. To support more than 65535 segments (program headers), Sun's "Linker and Libraries Guide" table 7-7 (or 12-7, depending on document version) prescribes a special first section header where sh_info represents the real number of program headers. Test code to follow, when it is ready. Reference: http://docs.oracle.com/cd/E18752_01/pdf/817-1984.pdf Reviewed by: emaste, markj Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D7255
2016-07-20 16:59:36 +00:00
phdr = (Elf_Phdr *)((char *)hdr + ehdr->e_phoff);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
/* The note segement. */
phdr->p_type = PT_NOTE;
phdr->p_offset = hdrsize;
phdr->p_vaddr = 0;
phdr->p_paddr = 0;
phdr->p_filesz = notesz;
phdr->p_memsz = 0;
phdr->p_flags = PF_R;
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
phdr->p_align = ELF_NOTE_ROUNDSIZE;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
phdr++;
/* All the writable segments from the program. */
phc.phdr = phdr;
phc.offset = round_page(hdrsize + notesz);
imgact_elf: Rename the segment iterator to match reality The each_writable_segment routine evaluates segments on a slightly little more nuanced metric than simply "writable" or not. Rename the function to more closely match its behavior (each_dumpable_segment). Suggested by: jhb Sponsored by: EMC / Isilon Storage Division
2016-07-20 22:51:33 +00:00
each_dumpable_segment(td, cb_put_phdr, &phc);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
}
static size_t
register_note(struct note_info_list *list, int type, outfunc_t out, void *arg)
{
struct note_info *ninfo;
size_t size, notesize;
size = 0;
out(arg, NULL, &size);
ninfo = malloc(sizeof(*ninfo), M_TEMP, M_ZERO | M_WAITOK);
ninfo->type = type;
ninfo->outfunc = out;
ninfo->outarg = arg;
ninfo->outsize = size;
TAILQ_INSERT_TAIL(list, ninfo, link);
if (type == -1)
return (size);
notesize = sizeof(Elf_Note) + /* note header */
Improve support for XSAVE with debuggers. - Dump an NT_X86_XSTATE note if XSAVE is in use. This note is designed to match what Linux does in that 1) it dumps the entire XSAVE area including the fxsave state, and 2) it stashes a copy of the current xsave mask in the unused padding between the fxsave state and the xstate header at the same location used by Linux. - Teach readelf() to recognize NT_X86_XSTATE notes. - Change PT_GET/SETXSTATE to take the entire XSAVE state instead of only the extra portion. This avoids having to always make two ptrace() calls to get or set the full XSAVE state. - Add a PT_GET_XSTATE_INFO which returns the length of the current XSTATE save area (so the size of the buffer needed for PT_GETXSTATE) and the current XSAVE mask (%xcr0). Differential Revision: https://reviews.freebsd.org/D1193 Reviewed by: kib MFC after: 2 weeks
2014-11-21 20:53:17 +00:00
roundup2(sizeof(FREEBSD_ABI_VENDOR), ELF_NOTE_ROUNDSIZE) +
/* note name */
roundup2(size, ELF_NOTE_ROUNDSIZE); /* note description */
return (notesize);
}
static size_t
append_note_data(const void *src, void *dst, size_t len)
{
size_t padded_len;
padded_len = roundup2(len, ELF_NOTE_ROUNDSIZE);
if (dst != NULL) {
bcopy(src, dst, len);
bzero((char *)dst + len, padded_len - len);
}
return (padded_len);
}
size_t
__elfN(populate_note)(int type, void *src, void *dst, size_t size, void **descp)
{
Elf_Note *note;
char *buf;
size_t notesize;
buf = dst;
if (buf != NULL) {
note = (Elf_Note *)buf;
note->n_namesz = sizeof(FREEBSD_ABI_VENDOR);
note->n_descsz = size;
note->n_type = type;
buf += sizeof(*note);
buf += append_note_data(FREEBSD_ABI_VENDOR, buf,
sizeof(FREEBSD_ABI_VENDOR));
append_note_data(src, buf, size);
if (descp != NULL)
*descp = buf;
}
notesize = sizeof(Elf_Note) + /* note header */
roundup2(sizeof(FREEBSD_ABI_VENDOR), ELF_NOTE_ROUNDSIZE) +
/* note name */
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
roundup2(size, ELF_NOTE_ROUNDSIZE); /* note description */
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
return (notesize);
}
static void
__elfN(putnote)(struct note_info *ninfo, struct sbuf *sb)
{
Elf_Note note;
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
ssize_t old_len, sect_len;
size_t new_len, descsz, i;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
if (ninfo->type == -1) {
ninfo->outfunc(ninfo->outarg, sb, &ninfo->outsize);
return;
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
Improve support for XSAVE with debuggers. - Dump an NT_X86_XSTATE note if XSAVE is in use. This note is designed to match what Linux does in that 1) it dumps the entire XSAVE area including the fxsave state, and 2) it stashes a copy of the current xsave mask in the unused padding between the fxsave state and the xstate header at the same location used by Linux. - Teach readelf() to recognize NT_X86_XSTATE notes. - Change PT_GET/SETXSTATE to take the entire XSAVE state instead of only the extra portion. This avoids having to always make two ptrace() calls to get or set the full XSAVE state. - Add a PT_GET_XSTATE_INFO which returns the length of the current XSTATE save area (so the size of the buffer needed for PT_GETXSTATE) and the current XSAVE mask (%xcr0). Differential Revision: https://reviews.freebsd.org/D1193 Reviewed by: kib MFC after: 2 weeks
2014-11-21 20:53:17 +00:00
note.n_namesz = sizeof(FREEBSD_ABI_VENDOR);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
note.n_descsz = ninfo->outsize;
note.n_type = ninfo->type;
sbuf_bcat(sb, &note, sizeof(note));
sbuf_start_section(sb, &old_len);
Improve support for XSAVE with debuggers. - Dump an NT_X86_XSTATE note if XSAVE is in use. This note is designed to match what Linux does in that 1) it dumps the entire XSAVE area including the fxsave state, and 2) it stashes a copy of the current xsave mask in the unused padding between the fxsave state and the xstate header at the same location used by Linux. - Teach readelf() to recognize NT_X86_XSTATE notes. - Change PT_GET/SETXSTATE to take the entire XSAVE state instead of only the extra portion. This avoids having to always make two ptrace() calls to get or set the full XSAVE state. - Add a PT_GET_XSTATE_INFO which returns the length of the current XSTATE save area (so the size of the buffer needed for PT_GETXSTATE) and the current XSAVE mask (%xcr0). Differential Revision: https://reviews.freebsd.org/D1193 Reviewed by: kib MFC after: 2 weeks
2014-11-21 20:53:17 +00:00
sbuf_bcat(sb, FREEBSD_ABI_VENDOR, sizeof(FREEBSD_ABI_VENDOR));
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
sbuf_end_section(sb, old_len, ELF_NOTE_ROUNDSIZE, 0);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
if (note.n_descsz == 0)
return;
sbuf_start_section(sb, &old_len);
ninfo->outfunc(ninfo->outarg, sb, &ninfo->outsize);
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
sect_len = sbuf_end_section(sb, old_len, ELF_NOTE_ROUNDSIZE, 0);
if (sect_len < 0)
return;
new_len = (size_t)sect_len;
descsz = roundup(note.n_descsz, ELF_NOTE_ROUNDSIZE);
if (new_len < descsz) {
/*
* It is expected that individual note emitters will correctly
* predict their expected output size and fill up to that size
* themselves, padding in a format-specific way if needed.
* However, in case they don't, just do it here with zeros.
*/
for (i = 0; i < descsz - new_len; i++)
sbuf_putc(sb, 0);
} else if (new_len > descsz) {
/*
* We can't always truncate sb -- we may have drained some
* of it already.
*/
KASSERT(new_len == descsz, ("%s: Note type %u changed as we "
"read it (%zu > %zu). Since it is longer than "
"expected, this coredump's notes are corrupt. THIS "
"IS A BUG in the note_procstat routine for type %u.\n",
__func__, (unsigned)note.n_type, new_len, descsz,
(unsigned)note.n_type));
}
Erk. Revert back to 1.31, dumping only data and stack to the core file, until I can solve a panic that has just cropped up.
1998-09-15 22:23:12 +00:00
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
/*
* Miscellaneous note out functions.
*/
Provide groundwork for 32-bit binary compatibility on non-x86 platforms, for upcoming 64-bit PowerPC and MIPS support. This renames the COMPAT_IA32 option to COMPAT_FREEBSD32, removes some IA32-specific code from MI parts of the kernel and enhances the freebsd32 compatibility code to support big-endian platforms. Reviewed by: kib, jhb
2010-03-11 14:49:06 +00:00
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
#include <compat/freebsd32/freebsd32.h>
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
#include <compat/freebsd32/freebsd32_signal.h>
Provide groundwork for 32-bit binary compatibility on non-x86 platforms, for upcoming 64-bit PowerPC and MIPS support. This renames the COMPAT_IA32 option to COMPAT_FREEBSD32, removes some IA32-specific code from MI parts of the kernel and enhances the freebsd32 compatibility code to support big-endian platforms. Reviewed by: kib, jhb
2010-03-11 14:49:06 +00:00
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
typedef struct prstatus32 elf_prstatus_t;
typedef struct prpsinfo32 elf_prpsinfo_t;
typedef struct fpreg32 elf_prfpregset_t;
typedef struct fpreg32 elf_fpregset_t;
typedef struct reg32 elf_gregset_t;
Add the ability for GDB to printout the thread name along with other thread specific informations. In order to do that, and in order to avoid KBI breakage with existing infrastructure the following semantic is implemented: - For live programs, a new member to the PT_LWPINFO is added (pl_tdname) - For cores, a new ELF note is added (NT_THRMISC) that can be used for storing thread specific, miscellaneous, informations. Right now it is just popluated with a thread name. GDB, then, retrieves the correct informations from the corefile via the BFD interface, as it groks the ELF notes and create appropriate pseudo-sections. Sponsored by: Sandvine Incorporated Tested by: gianni Discussed with: dim, kan, kib MFC after: 2 weeks
2010-11-22 14:42:13 +00:00
typedef struct thrmisc32 elf_thrmisc_t;
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
#define ELF_KERN_PROC_MASK KERN_PROC_MASK32
typedef struct kinfo_proc32 elf_kinfo_proc_t;
typedef uint32_t elf_ps_strings_t;
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
#else
typedef prstatus_t elf_prstatus_t;
typedef prpsinfo_t elf_prpsinfo_t;
typedef prfpregset_t elf_prfpregset_t;
typedef prfpregset_t elf_fpregset_t;
typedef gregset_t elf_gregset_t;
Add the ability for GDB to printout the thread name along with other thread specific informations. In order to do that, and in order to avoid KBI breakage with existing infrastructure the following semantic is implemented: - For live programs, a new member to the PT_LWPINFO is added (pl_tdname) - For cores, a new ELF note is added (NT_THRMISC) that can be used for storing thread specific, miscellaneous, informations. Right now it is just popluated with a thread name. GDB, then, retrieves the correct informations from the corefile via the BFD interface, as it groks the ELF notes and create appropriate pseudo-sections. Sponsored by: Sandvine Incorporated Tested by: gianni Discussed with: dim, kan, kib MFC after: 2 weeks
2010-11-22 14:42:13 +00:00
typedef thrmisc_t elf_thrmisc_t;
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
#define ELF_KERN_PROC_MASK 0
typedef struct kinfo_proc elf_kinfo_proc_t;
typedef vm_offset_t elf_ps_strings_t;
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
#endif
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
static void
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
__elfN(note_prpsinfo)(void *arg, struct sbuf *sb, size_t *sizep)
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
{
Include command line arguments in core dump process info. Fill in pr_psargs in the NT_PRSINFO ELF core dump note with command line arguments. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D7116
2016-07-14 23:20:05 +00:00
struct sbuf sbarg;
size_t len;
char *cp, *end;
Allocate TIDs in thread_init() and deallocate them in thread_fini(). The overhead of unconditionally allocating TIDs (and likewise, unconditionally deallocating them), is amortized across multiple thread creations by the way UMA makes it possible to have type-stable storage. Previously the cost was kept down by having threads created as part of a fork operation use the process' PID as the TID. While this had some nice properties, it also introduced complexity in the way TIDs were allocated. Most importantly, by using the type-stable storage that UMA gives us this was also unnecessary. This change affects how core dumps are created and in particular how the PRSTATUS notes are dumped. Since we don't have a thread with a TID equalling the PID, we now need a different way to preserve the old and previous behavior. We do this by having the given thread (i.e. the thread passed to the core dump code in td) dump it's state first and fill in pr_pid with the actual PID. All other threads will have pr_pid contain their TIDs. The upshot of all this is that the debugger will now likely select the right LWP (=TID) as the initial thread. Credits to: julian@ for spotting how we can utilize UMA. Thanks to: all who provided julian@ with test results.
2004-06-26 18:58:22 +00:00
struct proc *p;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
elf_prpsinfo_t *psinfo;
Include command line arguments in core dump process info. Fill in pr_psargs in the NT_PRSINFO ELF core dump note with command line arguments. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D7116
2016-07-14 23:20:05 +00:00
int error;
Create NT_PRSTATUS and NT_FPREGSET notes for each and every thread in the process. This is required for proper debugging of corefiles created by 1:1 or M:N threaded processes. Add an XXX comment where we should actually call a function that dumps MD specific notes. An example of a MD specific note is the NT_PRXFPREG note for SSE registers. Since BFD creates non-annotated pseudo-sections for the first PRSTATUS and FPREGSET notes (non-annotated in the sense that the name of the section does not contain the pid/tid), make sure those sections describe the initial thread of the process (i.e. the thread which tid equals the pid). This is not strictly necessary, but makes sure that tools that use the non-annotated section names will not change behaviour due to this change. The practical upshot of this all is that one can see the threads in the debugger when looking at a corefile. For 1:1 threading this means that *all* threads are visible.
2004-04-03 20:25:41 +00:00
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
p = (struct proc *)arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(*psinfo), ("invalid size"));
psinfo = malloc(sizeof(*psinfo), M_TEMP, M_ZERO | M_WAITOK);
Create NT_PRSTATUS and NT_FPREGSET notes for each and every thread in the process. This is required for proper debugging of corefiles created by 1:1 or M:N threaded processes. Add an XXX comment where we should actually call a function that dumps MD specific notes. An example of a MD specific note is the NT_PRXFPREG note for SSE registers. Since BFD creates non-annotated pseudo-sections for the first PRSTATUS and FPREGSET notes (non-annotated in the sense that the name of the section does not contain the pid/tid), make sure those sections describe the initial thread of the process (i.e. the thread which tid equals the pid). This is not strictly necessary, but makes sure that tools that use the non-annotated section names will not change behaviour due to this change. The practical upshot of this all is that one can see the threads in the debugger when looking at a corefile. For 1:1 threading this means that *all* threads are visible.
2004-04-03 20:25:41 +00:00
psinfo->pr_version = PRPSINFO_VERSION;
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
psinfo->pr_psinfosz = sizeof(elf_prpsinfo_t);
Go back to using the process command name (p_comm) for the file name and command line arguments stored in the note at the beginning of a core dump instead of the current thread name. Reviewed by: julian
2008-05-15 03:07:34 +00:00
strlcpy(psinfo->pr_fname, p->p_comm, sizeof(psinfo->pr_fname));
Include command line arguments in core dump process info. Fill in pr_psargs in the NT_PRSINFO ELF core dump note with command line arguments. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D7116
2016-07-14 23:20:05 +00:00
PROC_LOCK(p);
if (p->p_args != NULL) {
len = sizeof(psinfo->pr_psargs) - 1;
if (len > p->p_args->ar_length)
len = p->p_args->ar_length;
memcpy(psinfo->pr_psargs, p->p_args->ar_args, len);
PROC_UNLOCK(p);
error = 0;
} else {
_PHOLD(p);
PROC_UNLOCK(p);
sbuf_new(&sbarg, psinfo->pr_psargs,
sizeof(psinfo->pr_psargs), SBUF_FIXEDLEN);
error = proc_getargv(curthread, p, &sbarg);
PRELE(p);
if (sbuf_finish(&sbarg) == 0)
len = sbuf_len(&sbarg) - 1;
else
len = sizeof(psinfo->pr_psargs) - 1;
sbuf_delete(&sbarg);
}
if (error || len == 0)
strlcpy(psinfo->pr_psargs, p->p_comm,
sizeof(psinfo->pr_psargs));
else {
KASSERT(len < sizeof(psinfo->pr_psargs),
("len is too long: %zu vs %zu", len,
sizeof(psinfo->pr_psargs)));
cp = psinfo->pr_psargs;
end = cp + len - 1;
for (;;) {
cp = memchr(cp, '\0', end - cp);
if (cp == NULL)
break;
*cp = ' ';
}
}
Include process IDs in core dumps. When threads were added to the kernel, the pr_pid member of the NT_PRSTATUS note was repurposed to store LWP IDs instead of process IDs. However, the process ID was no longer recorded in core dumps. This change adds a pr_pid field to prpsinfo (NT_PRSINFO). Rather than bumping the prpsinfo version number, note parsers can use the note's payload size to determine if pr_pid is present. Reviewed by: kib, emaste (older version) MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D7117
2016-07-18 15:14:23 +00:00
psinfo->pr_pid = p->p_pid;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
sbuf_bcat(sb, psinfo, sizeof(*psinfo));
free(psinfo, M_TEMP);
Create NT_PRSTATUS and NT_FPREGSET notes for each and every thread in the process. This is required for proper debugging of corefiles created by 1:1 or M:N threaded processes. Add an XXX comment where we should actually call a function that dumps MD specific notes. An example of a MD specific note is the NT_PRXFPREG note for SSE registers. Since BFD creates non-annotated pseudo-sections for the first PRSTATUS and FPREGSET notes (non-annotated in the sense that the name of the section does not contain the pid/tid), make sure those sections describe the initial thread of the process (i.e. the thread which tid equals the pid). This is not strictly necessary, but makes sure that tools that use the non-annotated section names will not change behaviour due to this change. The practical upshot of this all is that one can see the threads in the debugger when looking at a corefile. For 1:1 threading this means that *all* threads are visible.
2004-04-03 20:25:41 +00:00
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
*sizep = sizeof(*psinfo);
}
Create NT_PRSTATUS and NT_FPREGSET notes for each and every thread in the process. This is required for proper debugging of corefiles created by 1:1 or M:N threaded processes. Add an XXX comment where we should actually call a function that dumps MD specific notes. An example of a MD specific note is the NT_PRXFPREG note for SSE registers. Since BFD creates non-annotated pseudo-sections for the first PRSTATUS and FPREGSET notes (non-annotated in the sense that the name of the section does not contain the pid/tid), make sure those sections describe the initial thread of the process (i.e. the thread which tid equals the pid). This is not strictly necessary, but makes sure that tools that use the non-annotated section names will not change behaviour due to this change. The practical upshot of this all is that one can see the threads in the debugger when looking at a corefile. For 1:1 threading this means that *all* threads are visible.
2004-04-03 20:25:41 +00:00
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
static void
__elfN(note_prstatus)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
elf_prstatus_t *status;
td = (struct thread *)arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(*status), ("invalid size"));
status = malloc(sizeof(*status), M_TEMP, M_ZERO | M_WAITOK);
status->pr_version = PRSTATUS_VERSION;
status->pr_statussz = sizeof(elf_prstatus_t);
status->pr_gregsetsz = sizeof(elf_gregset_t);
status->pr_fpregsetsz = sizeof(elf_fpregset_t);
status->pr_osreldate = osreldate;
status->pr_cursig = td->td_proc->p_sig;
status->pr_pid = td->td_tid;
Provide groundwork for 32-bit binary compatibility on non-x86 platforms, for upcoming 64-bit PowerPC and MIPS support. This renames the COMPAT_IA32 option to COMPAT_FREEBSD32, removes some IA32-specific code from MI parts of the kernel and enhances the freebsd32 compatibility code to support big-endian platforms. Reviewed by: kib, jhb
2010-03-11 14:49:06 +00:00
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
fill_regs32(td, &status->pr_reg);
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
#else
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
fill_regs(td, &status->pr_reg);
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
#endif
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
sbuf_bcat(sb, status, sizeof(*status));
free(status, M_TEMP);
Allocate TIDs in thread_init() and deallocate them in thread_fini(). The overhead of unconditionally allocating TIDs (and likewise, unconditionally deallocating them), is amortized across multiple thread creations by the way UMA makes it possible to have type-stable storage. Previously the cost was kept down by having threads created as part of a fork operation use the process' PID as the TID. While this had some nice properties, it also introduced complexity in the way TIDs were allocated. Most importantly, by using the type-stable storage that UMA gives us this was also unnecessary. This change affects how core dumps are created and in particular how the PRSTATUS notes are dumped. Since we don't have a thread with a TID equalling the PID, we now need a different way to preserve the old and previous behavior. We do this by having the given thread (i.e. the thread passed to the core dump code in td) dump it's state first and fill in pr_pid with the actual PID. All other threads will have pr_pid contain their TIDs. The upshot of all this is that the debugger will now likely select the right LWP (=TID) as the initial thread. Credits to: julian@ for spotting how we can utilize UMA. Thanks to: all who provided julian@ with test results.
2004-06-26 18:58:22 +00:00
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
*sizep = sizeof(*status);
}
Create NT_PRSTATUS and NT_FPREGSET notes for each and every thread in the process. This is required for proper debugging of corefiles created by 1:1 or M:N threaded processes. Add an XXX comment where we should actually call a function that dumps MD specific notes. An example of a MD specific note is the NT_PRXFPREG note for SSE registers. Since BFD creates non-annotated pseudo-sections for the first PRSTATUS and FPREGSET notes (non-annotated in the sense that the name of the section does not contain the pid/tid), make sure those sections describe the initial thread of the process (i.e. the thread which tid equals the pid). This is not strictly necessary, but makes sure that tools that use the non-annotated section names will not change behaviour due to this change. The practical upshot of this all is that one can see the threads in the debugger when looking at a corefile. For 1:1 threading this means that *all* threads are visible.
2004-04-03 20:25:41 +00:00
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
static void
__elfN(note_fpregset)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
elf_prfpregset_t *fpregset;
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
td = (struct thread *)arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(*fpregset), ("invalid size"));
fpregset = malloc(sizeof(*fpregset), M_TEMP, M_ZERO | M_WAITOK);
Provide groundwork for 32-bit binary compatibility on non-x86 platforms, for upcoming 64-bit PowerPC and MIPS support. This renames the COMPAT_IA32 option to COMPAT_FREEBSD32, removes some IA32-specific code from MI parts of the kernel and enhances the freebsd32 compatibility code to support big-endian platforms. Reviewed by: kib, jhb
2010-03-11 14:49:06 +00:00
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
fill_fpregs32(td, fpregset);
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
#else
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
fill_fpregs(td, fpregset);
Jumbo-commit to enhance 32 bit application support on 64 bit kernels. This is good enough to be able to run a RELENG_4 gdb binary against a RELENG_4 application, along with various other tools (eg: 4.x gcore). We use this at work. ia32_reg.[ch]: handle the 32 bit register file format, used by ptrace, procfs and core dumps. procfs_*regs.c: vary the format of proc/XXX/*regs depending on the client and target application. procfs_map.c: Don't print a 64 bit value to 32 bit consumers, or their sscanf fails. They expect an unsigned long. imgact_elf.c: produce a valid 32 bit coredump for 32 bit apps. sys_process.c: handle 32 bit consumers debugging 32 bit targets. Note that 64 bit consumers can still debug 32 bit targets. IA64 has got stubs for ia32_reg.c. Known limitations: a 5.x/6.x gdb uses get/setcontext(), which isn't implemented in the 32/64 wrapper yet. We also make a tiny patch to gdb pacify it over conflicting formats of ld-elf.so.1. Approved by: re
2005-06-30 07:49:22 +00:00
#endif
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
sbuf_bcat(sb, fpregset, sizeof(*fpregset));
free(fpregset, M_TEMP);
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
*sizep = sizeof(*fpregset);
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
}
static void
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
__elfN(note_thrmisc)(void *arg, struct sbuf *sb, size_t *sizep)
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
{
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
struct thread *td;
elf_thrmisc_t thrmisc;
td = (struct thread *)arg;
if (sb != NULL) {
KASSERT(*sizep == sizeof(thrmisc), ("invalid size"));
bzero(&thrmisc._pad, sizeof(thrmisc._pad));
strcpy(thrmisc.pr_tname, td->td_name);
sbuf_bcat(sb, &thrmisc, sizeof(thrmisc));
}
*sizep = sizeof(thrmisc);
}
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
static void
__elfN(note_ptlwpinfo)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
size_t size;
int structsize;
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
struct ptrace_lwpinfo32 pl;
#else
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
struct ptrace_lwpinfo pl;
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
#endif
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
td = (struct thread *)arg;
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
size = sizeof(structsize) + sizeof(pl);
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
structsize = sizeof(pl);
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
sbuf_bcat(sb, &structsize, sizeof(structsize));
bzero(&pl, sizeof(pl));
pl.pl_lwpid = td->td_tid;
pl.pl_event = PL_EVENT_NONE;
pl.pl_sigmask = td->td_sigmask;
pl.pl_siglist = td->td_siglist;
if (td->td_si.si_signo != 0) {
pl.pl_event = PL_EVENT_SIGNAL;
pl.pl_flags |= PL_FLAG_SI;
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
siginfo_to_siginfo32(&td->td_si, &pl.pl_siginfo);
#else
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
pl.pl_siginfo = td->td_si;
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
#endif
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
}
strcpy(pl.pl_tdname, td->td_name);
/* XXX TODO: supply more information in struct ptrace_lwpinfo*/
Store a 32-bit PT_LWPINFO struct for 32-bit process core dumps. Process core notes for a 32-bit process running on a 64-bit host need to use 32-bit structures so that the note layout matches the layout of notes of a core dump of a 32-bit process under a 32-bit kernel. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D11407
2017-06-29 21:31:13 +00:00
sbuf_bcat(sb, &pl, sizeof(pl));
Add support for capturing 'struct ptrace_lwpinfo' for signals resulting in a process dumping core in the corefile. Also extend procstat to view select members of 'struct ptrace_lwpinfo' from the contents of the note. Sponsored by: Dell EMC Isilon
2017-03-30 18:21:36 +00:00
}
*sizep = size;
}
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
/*
* Allow for MD specific notes, as well as any MD
* specific preparations for writing MI notes.
*/
static void
__elfN(note_threadmd)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct thread *td;
void *buf;
size_t size;
td = (struct thread *)arg;
size = *sizep;
if (size != 0 && sb != NULL)
buf = malloc(size, M_TEMP, M_ZERO | M_WAITOK);
Refine r264422: set buf to NULL only when we don't allocate memory, and free buf unconditionally. Requested by: kib MFC after: 1 week
2014-04-14 21:02:20 +00:00
else
buf = NULL;
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
size = 0;
__elfN(dump_thread)(td, buf, &size);
The size value should be asserted when it is known. Reported and tested by: pho Sponsored by: The FreeBSD Foundation
2014-11-22 18:15:02 +00:00
KASSERT(sb == NULL || *sizep == size, ("invalid size"));
Refine r264422: set buf to NULL only when we don't allocate memory, and free buf unconditionally. Requested by: kib MFC after: 1 week
2014-04-14 21:02:20 +00:00
if (size != 0 && sb != NULL)
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
sbuf_bcat(sb, buf, size);
Refine r264422: set buf to NULL only when we don't allocate memory, and free buf unconditionally. Requested by: kib MFC after: 1 week
2014-04-14 21:02:20 +00:00
free(buf, M_TEMP);
Re-factor coredump routines. For each type of notes an output function is provided, which is used either to calculate the note size or output it to sbuf. On the first pass the notes are registered in a list and the resulting size is found, on the second pass the list is traversed outputing notes to sbuf. For the sbuf a drain routine is provided that writes data to a core file. The main goal of the change is to make coredump to write notes directly to the core file, without preliminary preparing them all in a memory buffer. Storing notes in memory is not a problem for the current, rather small, set of notes we write to the core, but it may becomes an issue when we start to store procstat notes. Reviewed by: jhb (initial version), kib Discussed with: jhb, kib MFC after: 3 weeks
2013-04-14 19:59:38 +00:00
*sizep = size;
Viola! The kernel now generates standard ELF core dumps for ELF executables. Currently only data and stack are included in the core dumps. I am looking into adding the other (mmapped) writable segments as well.
1998-09-14 22:46:08 +00:00
}
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
#ifdef KINFO_PROC_SIZE
CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
#endif
static void
__elfN(note_procstat_proc)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = (struct proc *)arg;
size = sizeof(structsize) + p->p_numthreads *
sizeof(elf_kinfo_proc_t);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(elf_kinfo_proc_t);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
kern_proc_out(p, sb, ELF_KERN_PROC_MASK);
}
*sizep = size;
}
#ifdef KINFO_FILE_SIZE
CTASSERT(sizeof(struct kinfo_file) == KINFO_FILE_SIZE);
#endif
static void
note_procstat_files(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
size_t size, sect_sz, i;
ssize_t start_len, sect_len;
int structsize, filedesc_flags;
Follow-up to r287442: Move sysctl to compiled-once file Avoid duplicate sysctl nodes. Found by: tijl Approved by: markj (mentor) Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3586
2015-09-07 16:44:28 +00:00
if (coredump_pack_fileinfo)
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
filedesc_flags = KERN_FILEDESC_PACK_KINFO;
else
filedesc_flags = 0;
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
p = (struct proc *)arg;
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
structsize = sizeof(struct kinfo_file);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
if (sb == NULL) {
size = 0;
sb = sbuf_new(NULL, NULL, 128, SBUF_FIXEDLEN);
Optimize kern.geom.conf* sysctls. On large systems those sysctls may generate megabytes of output. Before this change sbuf(9) code was resizing buffer by 4KB each time many times, generating tons of TLB shootdowns. Unfortunately in this case existing sbuf_new_for_sysctl() mechanism, supposed to help with this issue, is not applicable, since all the sbuf writes are done in different kernel thread. This change improves situation in two ways: - on first sysctl call, not providing any output buffer, it sets special sbuf drain function, just counting the data and so not needing big buffer; - on second sysctl call it uses as initial buffer size value saved on previous call, so that in most cases there will be no reallocation, unless GEOM topology changed significantly. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-06-18 21:05:10 +00:00
sbuf_set_drain(sb, sbuf_count_drain, &size);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
kern_proc_filedesc_out(p, sb, -1, filedesc_flags);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
sbuf_finish(sb);
sbuf_delete(sb);
*sizep = size;
} else {
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
sbuf_start_section(sb, &start_len);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
Detect badly behaved coredump note helpers Coredump notes depend on being able to invoke dump routines twice; once in a dry-run mode to get the size of the note, and another to actually emit the note to the corefile. When a note helper emits a different length section the second time around than the length it requested the first time, the kernel produces a corrupt coredump. NT_PROCSTAT_FILES output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' fd table via vn_fullpath. As vnodes may move around during dump, this is racy. So: - Detect badly behaved notes in putnote() and pad underfilled notes. - Add a fail point, debug.fail_point.fill_kinfo_vnode__random_path to exercise the NT_PROCSTAT_FILES corruption. It simply picks random lengths to expand or truncate paths to in fo_fill_kinfo_vnode(). - Add a sysctl, kern.coredump_pack_fileinfo, to allow users to disable kinfo packing for PROCSTAT_FILES notes. This should avoid both FILES note corruption and truncation, even if filenames change, at the cost of about 1 kiB in padding bloat per open fd. Document the new sysctl in core.5. - Fix note_procstat_files to self-limit in the 2nd pass. Since sometimes this will result in a short write, pad up to our advertised size. This addresses note corruption, at the risk of sometimes truncating the last several fd info entries. - Fix NT_PROCSTAT_FILES consumers libutil and libprocstat to grok the zero padding. With suggestions from: bjk, jhb, kib, wblock Approved by: markj (mentor) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3548
2015-09-03 20:32:10 +00:00
kern_proc_filedesc_out(p, sb, *sizep - sizeof(structsize),
filedesc_flags);
sect_len = sbuf_end_section(sb, start_len, 0, 0);
if (sect_len < 0)
return;
sect_sz = sect_len;
KASSERT(sect_sz <= *sizep,
("kern_proc_filedesc_out did not respect maxlen; "
"requested %zu, got %zu", *sizep - sizeof(structsize),
sect_sz - sizeof(structsize)));
for (i = 0; i < *sizep - sect_sz && sb->s_error == 0; i++)
sbuf_putc(sb, 0);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
}
}
#ifdef KINFO_VMENTRY_SIZE
CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
#endif
static void
note_procstat_vmmap(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
Fix core corruption caused by race in note_procstat_vmmap This fix is spiritually similar to r287442 and was discovered thanks to the KASSERT added in that revision. NT_PROCSTAT_VMMAP output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' vm map via vn_fullpath. As vnodes may move during coredump, this is racy. We do not remove the race, only prevent it from causing coredump corruption. - Add a sysctl, kern.coredump_pack_vmmapinfo, to allow users to disable kinfo packing for PROCSTAT_VMMAP notes. This avoids VMMAP corruption and truncation, even if names change, at the cost of up to PATH_MAX bytes per mapped object. The new sysctl is documented in core.5. - Fix note_procstat_vmmap to self-limit in the second pass. This addresses corruption, at the cost of sometimes producing a truncated result. - Fix PROCSTAT_VMMAP consumers libutil (and libprocstat, via copy-paste) to grok the new zero padding. Reported by: pho (https://people.freebsd.org/~pho/stress/log/datamove4-2.txt) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3824
2015-10-06 18:07:00 +00:00
int structsize, vmmap_flags;
if (coredump_pack_vmmapinfo)
vmmap_flags = KERN_VMMAP_PACK_KINFO;
else
vmmap_flags = 0;
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
p = (struct proc *)arg;
Fix core corruption caused by race in note_procstat_vmmap This fix is spiritually similar to r287442 and was discovered thanks to the KASSERT added in that revision. NT_PROCSTAT_VMMAP output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' vm map via vn_fullpath. As vnodes may move during coredump, this is racy. We do not remove the race, only prevent it from causing coredump corruption. - Add a sysctl, kern.coredump_pack_vmmapinfo, to allow users to disable kinfo packing for PROCSTAT_VMMAP notes. This avoids VMMAP corruption and truncation, even if names change, at the cost of up to PATH_MAX bytes per mapped object. The new sysctl is documented in core.5. - Fix note_procstat_vmmap to self-limit in the second pass. This addresses corruption, at the cost of sometimes producing a truncated result. - Fix PROCSTAT_VMMAP consumers libutil (and libprocstat, via copy-paste) to grok the new zero padding. Reported by: pho (https://people.freebsd.org/~pho/stress/log/datamove4-2.txt) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3824
2015-10-06 18:07:00 +00:00
structsize = sizeof(struct kinfo_vmentry);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
if (sb == NULL) {
size = 0;
sb = sbuf_new(NULL, NULL, 128, SBUF_FIXEDLEN);
Optimize kern.geom.conf* sysctls. On large systems those sysctls may generate megabytes of output. Before this change sbuf(9) code was resizing buffer by 4KB each time many times, generating tons of TLB shootdowns. Unfortunately in this case existing sbuf_new_for_sysctl() mechanism, supposed to help with this issue, is not applicable, since all the sbuf writes are done in different kernel thread. This change improves situation in two ways: - on first sysctl call, not providing any output buffer, it sets special sbuf drain function, just counting the data and so not needing big buffer; - on second sysctl call it uses as initial buffer size value saved on previous call, so that in most cases there will be no reallocation, unless GEOM topology changed significantly. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-06-18 21:05:10 +00:00
sbuf_set_drain(sb, sbuf_count_drain, &size);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
Fix core corruption caused by race in note_procstat_vmmap This fix is spiritually similar to r287442 and was discovered thanks to the KASSERT added in that revision. NT_PROCSTAT_VMMAP output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' vm map via vn_fullpath. As vnodes may move during coredump, this is racy. We do not remove the race, only prevent it from causing coredump corruption. - Add a sysctl, kern.coredump_pack_vmmapinfo, to allow users to disable kinfo packing for PROCSTAT_VMMAP notes. This avoids VMMAP corruption and truncation, even if names change, at the cost of up to PATH_MAX bytes per mapped object. The new sysctl is documented in core.5. - Fix note_procstat_vmmap to self-limit in the second pass. This addresses corruption, at the cost of sometimes producing a truncated result. - Fix PROCSTAT_VMMAP consumers libutil (and libprocstat, via copy-paste) to grok the new zero padding. Reported by: pho (https://people.freebsd.org/~pho/stress/log/datamove4-2.txt) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3824
2015-10-06 18:07:00 +00:00
kern_proc_vmmap_out(p, sb, -1, vmmap_flags);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
sbuf_finish(sb);
sbuf_delete(sb);
*sizep = size;
} else {
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
Fix core corruption caused by race in note_procstat_vmmap This fix is spiritually similar to r287442 and was discovered thanks to the KASSERT added in that revision. NT_PROCSTAT_VMMAP output length, when packing kinfo structs, is tied to the length of filenames corresponding to vnodes in the process' vm map via vn_fullpath. As vnodes may move during coredump, this is racy. We do not remove the race, only prevent it from causing coredump corruption. - Add a sysctl, kern.coredump_pack_vmmapinfo, to allow users to disable kinfo packing for PROCSTAT_VMMAP notes. This avoids VMMAP corruption and truncation, even if names change, at the cost of up to PATH_MAX bytes per mapped object. The new sysctl is documented in core.5. - Fix note_procstat_vmmap to self-limit in the second pass. This addresses corruption, at the cost of sometimes producing a truncated result. - Fix PROCSTAT_VMMAP consumers libutil (and libprocstat, via copy-paste) to grok the new zero padding. Reported by: pho (https://people.freebsd.org/~pho/stress/log/datamove4-2.txt) Relnotes: yes Sponsored by: EMC / Isilon Storage Division Differential Revision: https://reviews.freebsd.org/D3824
2015-10-06 18:07:00 +00:00
kern_proc_vmmap_out(p, sb, *sizep - sizeof(structsize),
vmmap_flags);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
}
}
static void
note_procstat_groups(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = (struct proc *)arg;
size = sizeof(structsize) + p->p_ucred->cr_ngroups * sizeof(gid_t);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(gid_t);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, p->p_ucred->cr_groups, p->p_ucred->cr_ngroups *
sizeof(gid_t));
}
*sizep = size;
}
static void
note_procstat_umask(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = (struct proc *)arg;
size = sizeof(structsize) + sizeof(p->p_fd->fd_cmask);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(p->p_fd->fd_cmask);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, &p->p_fd->fd_cmask, sizeof(p->p_fd->fd_cmask));
}
*sizep = size;
}
static void
note_procstat_rlimit(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
struct rlimit rlim[RLIM_NLIMITS];
size_t size;
int structsize, i;
p = (struct proc *)arg;
size = sizeof(structsize) + sizeof(rlim);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(rlim);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PROC_LOCK(p);
for (i = 0; i < RLIM_NLIMITS; i++)
Implement lockless resource limits. Use the same scheme implemented to manage credentials. Code needing to look at process's credentials (as opposed to thred's) is provided with *_proc variants of relevant functions. Places which possibly had to take the proc lock anyway still use the proc pointer to access limits.
2015-06-10 10:48:12 +00:00
lim_rlimit_proc(p, i, &rlim[i]);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
PROC_UNLOCK(p);
sbuf_bcat(sb, rlim, sizeof(rlim));
}
*sizep = size;
}
static void
note_procstat_osrel(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = (struct proc *)arg;
size = sizeof(structsize) + sizeof(p->p_osrel);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(p->p_osrel);
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, &p->p_osrel, sizeof(p->p_osrel));
}
*sizep = size;
}
static void
__elfN(note_procstat_psstrings)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
elf_ps_strings_t ps_strings;
size_t size;
int structsize;
p = (struct proc *)arg;
size = sizeof(structsize) + sizeof(ps_strings);
if (sb != NULL) {
KASSERT(*sizep == size, ("invalid size"));
structsize = sizeof(ps_strings);
#if defined(COMPAT_FREEBSD32) && __ELF_WORD_SIZE == 32
ps_strings = PTROUT(p->p_sysent->sv_psstrings);
#else
ps_strings = p->p_sysent->sv_psstrings;
#endif
sbuf_bcat(sb, &structsize, sizeof(structsize));
sbuf_bcat(sb, &ps_strings, sizeof(ps_strings));
}
*sizep = size;
}
static void
__elfN(note_procstat_auxv)(void *arg, struct sbuf *sb, size_t *sizep)
{
struct proc *p;
size_t size;
int structsize;
p = (struct proc *)arg;
if (sb == NULL) {
size = 0;
sb = sbuf_new(NULL, NULL, 128, SBUF_FIXEDLEN);
Optimize kern.geom.conf* sysctls. On large systems those sysctls may generate megabytes of output. Before this change sbuf(9) code was resizing buffer by 4KB each time many times, generating tons of TLB shootdowns. Unfortunately in this case existing sbuf_new_for_sysctl() mechanism, supposed to help with this issue, is not applicable, since all the sbuf writes are done in different kernel thread. This change improves situation in two ways: - on first sysctl call, not providing any output buffer, it sets special sbuf drain function, just counting the data and so not needing big buffer; - on second sysctl call it uses as initial buffer size value saved on previous call, so that in most cases there will be no reallocation, unless GEOM topology changed significantly. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-06-18 21:05:10 +00:00
sbuf_set_drain(sb, sbuf_count_drain, &size);
Add a new set of notes to a process core dump to store procstat data. The notes format is a header of sizeof(int), which stores the size of the corresponding data structure to provide some versioning, and data in the format as it is returned by a related sysctl call. The userland tools (procstat(1)) will be taught to extract this data, providing additional info for postmortem analysis. PR: kern/173723 Suggested by: jhb Discussed with: jhb, kib Reviewed by: jhb (initial version), kib MFC after: 1 month
2013-04-16 19:19:14 +00:00
sbuf_bcat(sb, &structsize, sizeof(structsize));
PHOLD(p);
proc_getauxv(curthread, p, sb);
PRELE(p);
sbuf_finish(sb);
sbuf_delete(sb);
*sizep = size;
} else {
structsize = sizeof(Elf_Auxinfo);
sbuf_bcat(sb, &structsize, sizeof(structsize));
PHOLD(p);
proc_getauxv(curthread, p, sb);
PRELE(p);
}
}
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
static boolean_t
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
__elfN(parse_notes)(struct image_params *imgp, Elf_Note *checknote,
const char *note_vendor, const Elf_Phdr *pnote,
boolean_t (*cb)(const Elf_Note *, void *, boolean_t *), void *cb_arg)
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
{
Fix several issues with parsing the notes for ELF objects. Badly formed ELF note may cause the caclulated pointer to the next note to point both after the note region, that was checked in the code, but also to point before the region, that was not checked [1]. Remember the first note location in note0 and leap out if the note is not between note0 and note_end. In the similar way, badly formed note may cause infinite loop by pointing next note into the same or previous note. Guard against this by limiting amount of loop iterations by arbitrary choosen big number. For clarity, check the calculated note alignment in each iteration. Reported by: Chris Palmer <chris noncombatant org> [1] PR: kern/132886 Reviewed and tested by: dchagin MFC after: 3 days
2009-03-22 13:42:41 +00:00
const Elf_Note *note, *note0, *note_end;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
const char *note_name;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
char *buf;
int i, error;
boolean_t res;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
/* We need some limit, might as well use PAGE_SIZE. */
if (pnote == NULL || pnote->p_filesz > PAGE_SIZE)
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
return (FALSE);
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
ASSERT_VOP_LOCKED(imgp->vp, "parse_notes");
if (pnote->p_offset > PAGE_SIZE ||
pnote->p_filesz > PAGE_SIZE - pnote->p_offset) {
imgact_elf: do not relock the text vnode if possible. We unlock the vnode around malloc(M_WAITOK), to make it possible for pagedaemon to flush vnode pages for us. Instead of doing it unconditionally, first try M_NOWAIT allocation, which typically succeed. Only on failure, unlock the vnode and retry with M_WAITOK. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:04:01 +00:00
buf = malloc(pnote->p_filesz, M_TEMP, M_NOWAIT);
if (buf == NULL) {
VOP_UNLOCK(imgp->vp, 0);
buf = malloc(pnote->p_filesz, M_TEMP, M_WAITOK);
Switch to use shared vnode locks for text files during image activation. kern_execve() locks text vnode exclusive to be able to set and clear VV_TEXT flag. VV_TEXT is mutually exclusive with the v_writecount > 0 condition. The change removes VV_TEXT, replacing it with the condition v_writecount <= -1, and puts v_writecount under the vnode interlock. Each text reference decrements v_writecount. To clear the text reference when the segment is unmapped, it is recorded in the vm_map_entry backed by the text file as MAP_ENTRY_VN_TEXT flag, and v_writecount is incremented on the map entry removal The operations like VOP_ADD_WRITECOUNT() and VOP_SET_TEXT() check that v_writecount does not contradict the desired change. vn_writecheck() is now racy and its use was eliminated everywhere except access. Atomic check for writeability and increment of v_writecount is performed by the VOP. vn_truncate() now increments v_writecount around VOP_SETATTR() call, lack of which is arguably a bug on its own. nullfs bypasses v_writecount to the lower vnode always, so nullfs vnode has its own v_writecount correct, and lower vnode gets all references, since object->handle is always lower vnode. On the text vnode' vm object dealloc, the v_writecount value is reset to zero, and deadfs vop_unset_text short-circuit the operation. Reclamation of lowervp always reclaims all nullfs vnodes referencing lowervp first, so no stray references are left. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 month Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:20:43 +00:00
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
imgact_elf: do not relock the text vnode if possible. We unlock the vnode around malloc(M_WAITOK), to make it possible for pagedaemon to flush vnode pages for us. Instead of doing it unconditionally, first try M_NOWAIT allocation, which typically succeed. Only on failure, unlock the vnode and retry with M_WAITOK. Reviewed by: markj, trasz Tested by: mjg, pho Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D19923
2019-05-05 11:04:01 +00:00
}
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
error = vn_rdwr(UIO_READ, imgp->vp, buf, pnote->p_filesz,
pnote->p_offset, UIO_SYSSPACE, IO_NODELOCKED,
curthread->td_ucred, NOCRED, NULL, curthread);
if (error != 0) {
uprintf("i/o error PT_NOTE\n");
Trivial reduction of the code duplication, reuse the return FALSE code. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:16:01 +00:00
goto retf;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
}
note = note0 = (const Elf_Note *)buf;
note_end = (const Elf_Note *)(buf + pnote->p_filesz);
} else {
note = note0 = (const Elf_Note *)(imgp->image_header +
pnote->p_offset);
note_end = (const Elf_Note *)(imgp->image_header +
pnote->p_offset + pnote->p_filesz);
buf = NULL;
}
Fix several issues with parsing the notes for ELF objects. Badly formed ELF note may cause the caclulated pointer to the next note to point both after the note region, that was checked in the code, but also to point before the region, that was not checked [1]. Remember the first note location in note0 and leap out if the note is not between note0 and note_end. In the similar way, badly formed note may cause infinite loop by pointing next note into the same or previous note. Guard against this by limiting amount of loop iterations by arbitrary choosen big number. For clarity, check the calculated note alignment in each iteration. Reported by: Chris Palmer <chris noncombatant org> [1] PR: kern/132886 Reviewed and tested by: dchagin MFC after: 3 days
2009-03-22 13:42:41 +00:00
for (i = 0; i < 100 && note >= note0 && note < note_end; i++) {
Fix several reads beyond the mapped first page of the binary in the ELF parser. Specifically, do not allow note reader and interpreter path comparision in the brandelf code to read past end of the page. This may happen if specially crafter ELF image is activated. Submitted by: Lukasz Wojcik <lukasz.wojcik zoho com> MFC after: 3 days
2012-07-19 11:15:53 +00:00
if (!aligned(note, Elf32_Addr) || (const char *)note_end -
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
(const char *)note < sizeof(Elf_Note)) {
Trivial reduction of the code duplication, reuse the return FALSE code. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:16:01 +00:00
goto retf;
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
}
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
if (note->n_namesz != checknote->n_namesz ||
note->n_descsz != checknote->n_descsz ||
note->n_type != checknote->n_type)
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
goto nextnote;
note_name = (const char *)(note + 1);
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
if (note_name + checknote->n_namesz >=
(const char *)note_end || strncmp(note_vendor,
note_name, checknote->n_namesz) != 0)
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
goto nextnote;
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
if (cb(note, cb_arg, &res))
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
goto ret;
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
nextnote:
note = (const Elf_Note *)((const char *)(note + 1) +
Introduce a constant, ELF_NOTE_ROUNDSIZE, which evidently declare our intention to use 4-byte padding for elf notes. MFC after: 3 weeks
2013-05-01 14:59:16 +00:00
roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE) +
roundup2(note->n_descsz, ELF_NOTE_ROUNDSIZE));
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
}
Trivial reduction of the code duplication, reuse the return FALSE code. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:16:01 +00:00
retf:
Allow PT_INTERP and PT_NOTES segments to be located anywhere in the executable image. Keep one page (arbitrary) limit on the max allowed size of the PT_NOTES. The ELF image activators still require that program headers of the executable are fully contained in the first page of the image file. Reviewed by: emaste, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential revision: https://reviews.freebsd.org/D3871
2015-10-14 18:27:35 +00:00
res = FALSE;
ret:
free(buf, M_TEMP);
return (res);
Implement new way of branding ELF binaries by looking to a ".note.ABI-tag" section. The search order of a brand is changed, now first of all the ".note.ABI-tag" is looked through. Move code which fetch osreldate for ELF binary to check_note() handler. PR: 118473 Approved by: kib (mentor)
2009-03-13 16:40:51 +00:00
}
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
struct brandnote_cb_arg {
Elf_Brandnote *brandnote;
int32_t *osrel;
};
static boolean_t
brandnote_cb(const Elf_Note *note, void *arg0, boolean_t *res)
{
struct brandnote_cb_arg *arg;
arg = arg0;
/*
* Fetch the osreldate for binary from the ELF OSABI-note if
* necessary.
*/
*res = (arg->brandnote->flags & BN_TRANSLATE_OSREL) != 0 &&
arg->brandnote->trans_osrel != NULL ?
arg->brandnote->trans_osrel(note, arg->osrel) : TRUE;
return (TRUE);
}
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
static Elf_Note fctl_note = {
.n_namesz = sizeof(FREEBSD_ABI_VENDOR),
.n_descsz = sizeof(uint32_t),
.n_type = NT_FREEBSD_FEATURE_CTL,
};
struct fctl_cb_arg {
uint32_t *fctl0;
};
static boolean_t
note_fctl_cb(const Elf_Note *note, void *arg0, boolean_t *res)
{
struct fctl_cb_arg *arg;
const Elf32_Word *desc;
uintptr_t p;
arg = arg0;
p = (uintptr_t)(note + 1);
p += roundup2(note->n_namesz, ELF_NOTE_ROUNDSIZE);
desc = (const Elf32_Word *)p;
*arg->fctl0 = desc[0];
return (TRUE);
}
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
/*
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
* Try to find the appropriate ABI-note section for checknote, fetch
* the osreldate and feature control flags for binary from the ELF
* OSABI-note. Only the first page of the image is searched, the same
* as for headers.
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
*/
static boolean_t
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
__elfN(check_note)(struct image_params *imgp, Elf_Brandnote *brandnote,
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
int32_t *osrel, uint32_t *fctl0)
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
{
const Elf_Phdr *phdr;
const Elf_Ehdr *hdr;
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
struct brandnote_cb_arg b_arg;
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
struct fctl_cb_arg f_arg;
int i, j;
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
hdr = (const Elf_Ehdr *)imgp->image_header;
phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff);
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
b_arg.brandnote = brandnote;
b_arg.osrel = osrel;
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
f_arg.fctl0 = fctl0;
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
for (i = 0; i < hdr->e_phnum; i++) {
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
if (phdr[i].p_type == PT_NOTE && __elfN(parse_notes)(imgp,
&brandnote->hdr, brandnote->vendor, &phdr[i], brandnote_cb,
&b_arg)) {
Parse FreeBSD Feature Control note on the ELF image activation. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:33:55 +00:00
for (j = 0; j < hdr->e_phnum; j++) {
if (phdr[j].p_type == PT_NOTE &&
__elfN(parse_notes)(imgp, &fctl_note,
FREEBSD_ABI_VENDOR, &phdr[j],
note_fctl_cb, &f_arg))
break;
}
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
return (TRUE);
Generalize ELF parse_notes(). Remove the knowledge of the ABI note type and brandnote from it, instead provide it with a callback to do note-specific matching and data fetching. Implement callback to match against ELF brand. Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2018-11-23 23:29:14 +00:00
}
ELF image can have several PT_NOTE program headers. Look for the ELF brand note in each header, instead of using only first one. Reviewed by: kan Tested by: andrew (arm), flo (sparc64) MFC after: 3 weeks
2012-03-11 19:38:49 +00:00
}
return (FALSE);
}
First attempt at FreeBSD & Linux ELF support. Compile and link a new kernel, that will give native ELF support, and provide the hooks for other ELF interpreters as well. To make native ELF binaries use John Polstras elf-kit-1.0.1.. For the time being also use his ld-elf.so.1 and put it in /usr/libexec. The Linux emulator has been enhanced to also run ELF binaries, it is however in its very first incarnation. Just get some Linux ELF libs (Slackware-3.0) and put them in the prober place (/compat/linux/...). I've ben able to run all the Slackware-3.0 binaries I've tried so far. (No it won't run quake yet :)
1996-03-10 08:42:54 +00:00
/*
* Tell kern_execve.c about it, with a little help from the linker.
*/
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
static struct execsw __elfN(execsw) = {
Use C99 designated initializers for struct execsw It it makes use slightly more clear and facilitates grepping.
2018-03-13 13:09:10 +00:00
.ex_imgact = __CONCAT(exec_, __elfN(imgact)),
.ex_name = __XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
Improve the way that an elf image activator for an alternate word size is included in the kernel. Include imgact_elf.c in conf/files, instead of both imgact_elf32.c and imgact_elf64.c, which will use the default word size for an architecture as defined in machine/elf.h. Architectures that wish to build an additional image activator for an alternate word size can include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which allows it to be dependent on MD options instead of solely on architecture. Glanced at by: peter
2003-01-04 22:07:48 +00:00
};
EXEC_SET(__CONCAT(elf, __ELF_WORD_SIZE), __elfN(execsw));
Merge projects/enhanced_coredumps (r204346) into HEAD: Enhanced process coredump routines. This brings in the following features: 1) Limit number of cores per process via the %I coredump formatter. Example: if corefilename is set to %N.%I.core AND num_cores = 3, then if a process "rpd" cores, then the corefile will be named "rpd.0.core", however if it cores again, then the kernel will generate "rpd.1.core" until we hit the limit of "num_cores". this is useful to get several corefiles, but also prevent filling the machine with corefiles. 2) Encode machine hostname in core dump name via %H. 3) Compress coredumps, useful for embedded platforms with limited space. A sysctl kern.compress_user_cores is made available if turned on. To enable compressed coredumps, the following config options need to be set: options COMPRESS_USER_CORES device zlib # brings in the zlib requirements. device gzio # brings in the kernel vnode gzip output module. 4) Eventhandlers are fired to indicate coredumps in progress. 5) The imgact sv_coredump routine has grown a flag to pass in more state, currently this is used only for passing a flag down to compress the coredump or not. Note that the gzio facility can be used for generic output of gzip'd streams via vnodes. Obtained from: Juniper Networks Reviewed by: kan
2010-03-02 06:58:58 +00:00
Collect code to translate between vm_prot_t and p_flags into helper functions. MFC after: 1 week
2011-01-08 16:02:14 +00:00
static vm_prot_t
__elfN(trans_prot)(Elf_Word flags)
{
vm_prot_t prot;
prot = 0;
if (flags & PF_X)
prot |= VM_PROT_EXECUTE;
if (flags & PF_W)
prot |= VM_PROT_WRITE;
if (flags & PF_R)
prot |= VM_PROT_READ;
Port sysctl kern.elf32.read_exec from amd64 to i386. Make it more comprehensive on i386, by not setting nx bit for any mapping, not just adding PF_X to all kernel-loaded ELF segments. This is needed for the compatibility with older i386 programs that assume that read access implies exec, e.g. old X servers with hand-rolled module loader. Reported and tested by: bde Sponsored by: The FreeBSD Foundation MFC after: 1 week
2019-02-07 02:17:34 +00:00
#if __ELF_WORD_SIZE == 32 && (defined(__amd64__) || defined(__i386__))
Control the execution permission of the readable segments for i386 binaries on the amd64 and ia64 with the sysctl, instead of unconditionally enabling it. Reviewed by: marcel
2011-10-15 12:35:18 +00:00
if (i386_read_exec && (flags & PF_R))
In elf32_trans_prot() and when compiling for amd64 or ia64, add PROT_EXECUTE when PROT_READ is needed. By default i386 allows execution when reading is allowed and JDK 1.4.x depends on that.
2011-10-13 16:16:46 +00:00
prot |= VM_PROT_EXECUTE;
#endif
Collect code to translate between vm_prot_t and p_flags into helper functions. MFC after: 1 week
2011-01-08 16:02:14 +00:00
return (prot);
}
static Elf_Word
__elfN(untrans_prot)(vm_prot_t prot)
{
Elf_Word flags;
flags = 0;
if (prot & VM_PROT_EXECUTE)
flags |= PF_X;
if (prot & VM_PROT_READ)
flags |= PF_R;
if (prot & VM_PROT_WRITE)
flags |= PF_W;
return (flags);
}
Make randomized stack gap between strings and pointers to argv/envs. This effectively makes the stack base on the csu _start entry randomized. The gap is enabled if ASLR is for the ABI is enabled, and then kern.elf{64,32}.aslr.stack_gap specify the max percentage of the initial stack size that can be wasted for gap. Setting it to zero disables the gap, and max is capped at 50%. Only amd64 for now. Reviewed by: cem, markj Discussed with: emaste MFC after: 2 weeks Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D21081
2019-07-31 20:23:10 +00:00
void
__elfN(stackgap)(struct image_params *imgp, u_long *stack_base)
{
u_long range, rbase, gap;
int pct;
if ((imgp->map_flags & MAP_ASLR) == 0)
return;
pct = __elfN(aslr_stack_gap);
if (pct == 0)
return;
if (pct > 50)
pct = 50;
range = imgp->eff_stack_sz * pct / 100;
arc4rand(&rbase, sizeof(rbase), 0);
gap = rbase % range;
gap &= ~(sizeof(u_long) - 1);
*stack_base -= gap;
}
Reference in New Issue Copy Permalink