Changelog
crash (7.0.7-1ubuntu1) utopic; urgency=medium
* Merge with Debian; remainging changes:
- Build for armhf, arm64.
- Minor fixes for live autopkgtest.
- debian/tests/live: Redirect gpg stderr to stdout, as that's the only
known and expected stderr source.
- Add ppc64el support.
crash (7.0.7-1) unstable; urgency=low
* Export the static ELF and compressed kdump vmcoreinfo_read_string()
functions from netdump.c and kdump.c via a new read_vmcoreinfo() method in
the global program_context structure. The function
get_log_from_vmcoreinfo() will access vmcoreinfo data via the new pointer
instead of requiring its callers to pass pointers to their
dumpfile-specific function.
* Linux 3.15 and later kernels configured with CONFIG_RANDOMIZE_BASE can be
now be readily identified because of new kernel symbols that have been
added. For those kernels, the new "--kaslr=<offset>" and/or
"--kaslr=auto" options are not necessary for ELF or compressed kdump
vmcores, or for live systems that have /proc/kallsyms showing the
relocated symbol values. A new KASLR initialization function called
kaslr_init() is now called by symtab_init() prior to the initial
symbol-sorting operation. If kaslr_init() determines that KASLR may be in
effect, it will trigger a search for the relevant vmlinux symbols during
the sorting operation, which in turn will cause the relocation value to be
automatically calculated.
* Implemented a new "bt -c cpu(s)" option to display the backtrace of the
active task on one or more cpus. The cpus must be specified in a comma-
and/or dash-separated list; for examples ""3", "1,8,9", "1-23", or
"1,8,9-14". Similar to "bt -a", the option is only applicable with crash
dumps.
* Fix for Linux 3.11 and later ARM kernels, in which all non-panicking cpus
offline themselves during a kdump procedure. This causes an invalid cpu
count determination during crash session initialization from an ARM
vmcore. The patch utilizes the cpu count found in the cpu_active_map if
it is greater than the count in the cpu_online_map. In addition, the
maximum NR_CPUS value for the ARM architecture has been raised from 4 to
32.
* Fix for the X86_64 "bt" command on Linux 3.3 and later kernels to properly
display exception frame register contents on NMI stacks. Kernel commit
3f3c8b8c4b2a34776c3470142a7c8baafcda6eb0 added 12 more values to the NMI
exception stack to handle nested NMIs caused by page faults or breakpoints
that could occur while handling an NMI exception.
* Kernel commit 28696f434fef0efa97534b59986ad33b9c4df7f8 changed the stack
layout again, swapping the location of the "saved" and "copied" registers.
This can be detected automatically, because the "copied" registers contain
either a copy of the "saved" registers, or point to "repeat_nmi". So, if
"repeat_nmi" is found as the return address, assume that this is the old
layout, and adjust the stack pointer again. Without the patch, incorrect
register values are displayed in the exception frame dump in the NMI stack
backtrace.
* Fix for the built-in "g" alias, which apparently has not worked correctly
since crash-5.1.4. Without the patch, if the "g" alias and the first
argument are separated by one space, then the first first character of
that argument would get stripped prior to being passed to the embedded gdb
module.
* Removed the BASELEVEL_REVISION string from defs.h, which serves no purpose
since the deprecation of the remote daemon, and typically has been out of
sync with the crash version.
* Fix for the "p", "irq", "struct", "union" and "*" commands if a cpu
specification contains an invalid cpu number. Without the patch, a
segmentation violation may be generated.
* Implemented a new capability for the "ptov" command that takes a per-cpu
offset and cpu specification argument and translates it into the kernel
virtual addresses for the cpus specified.
* Implemented a new "ps -m" option that is a similar, complementary option
to "ps -l", but which translates the task timestamp value from a decimal
or hexadecimal nanoseconds value into a more human-readable string
consisting of the number of days, hours, minutes, seconds and milliseconds
that have elapsed since the task started executing on a cpu. More
accurately described, it is the time difference between the timestamp
copied from the per-cpu runqueue clock when the task last started
executing compared to the most current value of the per-cpu runqueue
clock.
* In addition, a new "ps -C <cpu-specifier>" option has been added that can
only be used with "ps -l" and "ps -m", which sorts the global task list
into per-cpu blocks; the cpu-specifier uses the standard comma or dash
separated list, expressed as "-C 1,3,5", "-C 1-3", "-C 1,3,5-7,10", or
"-Call" or "-Ca" for all cpus.
* Implemented a new "runq -m" option that is a simliar, complementary option
to "runq -t", but which displays the amount of time that the active task
on each cpu has been running, expressed in a format consisting of days,
hours, minutes, seconds and milliseconds.
* Implemented a new "kmem -h" option that displays the address of each
hugepage hstate array entry, its hugepage size, its free and total counts,
and name string.
* Implemented a new "ps -S" option that displays a summary consisting of the
number of tasks in a task state.
* Fix for the "arguments-input-file" feature to protect against a called
command modifying an argument string. For example, the "struct" command
modifies "-l struct_name.member" argument strings, and so without the
patch, all iterative calls after the first one will fail.
* Fix failure to build from source when compiling the crash utility with
gcc-4.9 (closes: #746835)
* Fix for displaying enum values that are greater than 32-bits in size.
Without the patch, the upper 32-bits are clipped off and displayed as
integer-sized value.
* If the kernel (live or dumpfile) has the "kpatch" module installed, the
tag "[KPATCH]" will be displayed next to the kernel name in the initial
system banner and by the "sys" command.
* Fix for the "DEBUG KERNEL:" display in the initial system banner and by
the "sys" command when using a System.map file with a Linux 3.0 and later
debug kernel. Without the patch, the kernel version is not displayed in
parentheses following the debug kernel name.
* If the gdb-<version>.patch file has changed and a rebuild is being done
from within a previously-existing build tree, "patch -N" the gdb sources,
and start the rebuild from the gdb-<version> directory instead of the
gdb-<version>/gdb directory.
* Fix to prevent a possible segmentation violation generated by the "runq
-g" command when run on a very active live system due to an active task on
a cpu exiting while the command is running.
* Fix for the "runq -g" command on Linux 3.15 and later kernels, where the
cgroup_name() function now utilizes kernfs_name(). Without the patch, the
command fails with the error message "runq: invalid structure member
offset: cgroup_dentry".
* Fix for the "extend" command when running with an x86_64 crash binary that
was built with "make target=ARM64" in order to analyze ARM64 dumpfiles on
an x86_64 host. Without the patch, if the extend command is used with an
extension module built in the same manner, it fails with the message
"extend: <module>.so: not an ELF format object file".
* Introduce support for 32-bit ARM kernels that are configured with
CONFIG_ARM_LPAE. The patch implements the virtual-to-physical address
translation of 64-bit PTEs used by ARM LPAE kernels.
crash (7.0.6-1) unstable; urgency=low
* Fix for custom X86_64 kernels that change the declaration of the
context_switch() function so that it is not an inline function. Without
the patch, the message "crash: cannot determine thread return address" is
displayed during invocation, and backtraces of blocked tasks may have
missing or invalid frames.
* Fix to prevent a possible invocation-time error on Linux 3.7 and later
kernels configured with CONFIG_SLAB, running against vmcore files filtered
with the makedumpfile(8) facility. Without the patch, the message "crash:
page excluded: kernel virtual address: <address> type: kmem_cache buffer"
is immediately followed by the message "crash: unable to initialize kmem
slab cache subsystem". Because of a kernel data structure name change
from "cache_cache" to "kmem_cache_boot", the crash utility failed to
properly downsize the stored size of the kernel's kmem_cache data
structure from the size indicated by the vmlinux debuginfo data. This in
turn could lead to reading beyond the end of a kmem_cache data structure
into a page of memory that had been excluded from the vmcore. The fix was
also applied to kernels configured with CONFIG_SLUB.
* Added a new "--kaslr <offset>" command line option for X86_64 kernels that
are configured with CONFIG_RANDOMIZE_BASE. The offset value must be equal
to the difference between the symbol values compiled into the vmlinux file
and their relocated KASLR values.
* Added a new "--kaslr=auto" command line option for X86_64 kernels that
that are configured with CONFIG_RANDOMIZE_BASE. When set to "auto", the
KASLR relocation value will be determined automatically by comparing the
"_stext" symbol value compiled into the vmlinux file with the _stext
symbol value stored in kdump vmcoreinfo data; on live systems the
comparison will be made with the "_stext" symbol value that is found in
/proc/kallsyms.
* Enable kernel text line number capability for the "dis -l", "bt -l", "sys
-c", and "sym" commands for kernels that are configured with
CONFIG_RANDOMIZE_BASE.
* Fix for the "crash --log vmcore" command to account for the kernel data
structure and VMCOREINFO string name changes from "log" to "printk_log" in
Linux 3.11-rc4 and later kernels. Without the patch, the command fails
with the error message "crash: VMCOREINFO: no log buffer data".
* Adjustment to the internal symbol-handling to prevent the usage of kernel
system call alias/wrapper names, for examples, "SyS_read" and
"compat_SyS_futex" instead of "sys_read" and "compat_sys_futex". Without
the patch, commands such as "dis", "sym <address>", and "sys -c" display
the alias/wrapper name instead of the real system call name in Linux 3.10
and later kernels.
* Increase the internal hash queue head count from 128 to 32768. The hash
queue is used for gathering and verifying lists, and the original count of
128 may be overwhelmed if a list is extremely large. For example, on a
256GB system with 192GB of free pages, the "kmem -f" command takes hours
to complete; with this patch, the time is reduced to a few minutes. In
addition, a new command line option "--hash <count>" has been added to
allow a user to override the default hash queue head count of 32768.
* Fix for the "kmem -F" display and the "kmem -f <address>" or "kmem
<address>" options. Without the patch, "kmem -F" does not display the
first page in a list of free page blocks on its own line, but rather at
the end of the previous line that shows the area number, block size, and
free_area struct address that the page is linked to. Due to this error,
both "kmem -f <address>" and "kmem -f address>" would not find the
associated page or page block if it happened to be the first page or page
block in the list.
* Created a new feature for the internal do_list() function if it is
necessary to immediately perform a function for each entry in a list while
the list is being traversed. A callback function, and an option callback
data pointer, can be registered in the list_data structure. The address
of each entry in the list along with the optional callback data pointer
will be passed to the callback function. If desired, the callback
function may also dictate that do_list() should stop the list traversal
and return immediately to its caller.
* Made the "kmem -f <address>" and "kmem <address>" options more efficient
by using the new do_list() callback function feature above as well as
restricting the search to only the NUMA node that contains the address.
* If the first assembly language instruction in an X86_64 function is "nopl
0x0(%rax,%rax,1)" or "data32 data32 data32 xchg %ax,%ax", which are
generated when the ftrace facility is configured, the X86_64 "dis" command
will append "[FTRACE NOP]" to the line.
* Correction for the "crash -h" and crash.8 man page documentation of the
"--machdep phys_base=<physical-address>" command line option. In both
places the parameter mistakenly indicated "physbase".
* If a host build system does not have /usr/bin/wget installed, and the
crash package is built from a directory that was git-cloned from
github.com/crash-utility/crash.git, the error message has been clarified
to indicate "/usr/bin/wget is required to download gdb-7.6.tar.gz".
Without the patch, the message indicates "tar (child): gdb-7.6.tar.gz:
Cannot open: No such file or directory".
* Updated the ARM64 implementation to support Linux 3.13 and later kernels
that expand to a 42-bit address space when 64K pages are configured. This
is also the first crash version that has been tested on a live ARM64
system with 4K pages, where it cleanly make it to the "crash>" prompt.
However, it should be noted that some commands (most notably "bt") still
do not work as of yet.
* Document the "--machdep phys_offset=<physical-address>" command line
option for the ARM64 architecture in the crash.8 man page and the "crash
-h" output.
* Fix for KVM dumpfiles created with "virsh dump --memory-only" if an X86_64
kernel was loaded with a non-zero "phys_base". Without the patch, the
crash session fails with the warning message "WARNING: cannot read
linux_banner string" followed by the fatal error message "crash: vmlinux
and <dumpfile name> do not match!".
* Initial working implementation of the basic ARM64 "bt" command, with
several command options still under development. In-kernel exception
frames are only dumped if the exception handler function is contained
within the symbol boundaries from "__exception_text_start" to
"__exception_text_end"; when ARM64 kdump is eventually implemented,
further exception-related work will be resumed.
* Cleaned up the exception frame displays of 64-bit in-kernel and both
32-bit and 64-bit user-mode exceptions.
* Implemented support for the ARM64 "bt -e" option.
* Implemented support for the ARM64 "bt -l" option.
* Update for the X86_64 "bt -l" option such that it also displays the
available file and line number information for functions indicated as the
"exception RIP" in kernel exception frames. The line number information
will follow the exception frame register dump.
* Fix for the ARM64 virtual-to-physical translation of vmemmap page
structure addresses for kernels configured with 4K pages. Without the
patch, any command that required the contents of a page structure would
fail with a readmem error.
* Added support for the ARM64 architecture in the extensions/snap.c
extension module. Also fixed the progress percentage display to correct
for systems which have non-zero starting physical addresses.
* Implemented support for the ARM64 "bt -f" and "bt -F[F]" options.
* Increase the ARM64 PTRS_PER_PGD_L2_64K from 1024 to 9182 to account for
the Linux 3.13 increase of the ARM64 virtual address space size from 39 to
42 bits when 64K pages are configured. Without the patch, the warning
message "WARNING: cannot access vmalloc'd module memory" is displayed
during session initialization.
* Fix for the "vm -p" option on ARM64 so that file-backed pages are properly
translated to the filename and offset. Without the patch, file-backed
pages are erroneously shown as being backed on a swap device.
* Increment maximum ARM64 physical address from 40 to 48 bits to match
upstream kernel commit 87366d8cf7b3f6dc34633938aa8766e5a390ce33.
* Fix for a segmentation violation generated by the "crash -g vmlinux"
command on ARM64.
* Fix for the ARM64 "vtop <address>" command on kernels configured with 64K
pages if the address argument is located in the kernel logical memory map
region, which uses 512MB hugepage mappings. Without the patch, the
verbose page table walk mistakenly continues to the PTE level.
* Fix for ARM64 /proc/kcore support. Without the patch, the crash session
fails with the warning message "WARNING: cannot read linux_banner string"
followed by the fatal error message "crash: vmlinux and <dumpfile name> do
not match!". At this point in time, the kernel requires a patch to the
ARM64 kern_addr_valid() function to properly allow memory to be read from
the kernel logical memory map region.
-- Matthias Klose <email address hidden> Fri, 08 Aug 2014 14:03:57 +0200
