Since memory chunks can overlap for vr-kernel dumps stored in elf
format (ngdump, zfcpdump), we need to consider this when converting to
s390 dump output format. For that sort DFI memory chunks by start
address and adjust dfo_s390 logic for identifying memory gaps.
Otherwise we might end up with bogus DFO memory chunks being created.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
For vr-kernel dumps use the offsets stored in os_info entries for
virt to phys address conversion when dump virtual address is to be
read, (e.g. vmcoreinfo symbols) using similar method as implemented
in crash-utility.
It is mainly required for reading "init_uts_ns" symbol and, in case
of crashed kdump, "lowcore_ptr" symbol along with a pointers to the
lowcore of every CPU.
Acked-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Update verbose 'zgetdump -i' output with zlib info (internal zlib version
and zlib compression unit size).
The following new entriees are to be dispalyed:
Zlib version.......: 1
Zlib compression unit: 1 MB
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add support for vmdump 64big format to the zgetdump utility.
The result is an elf file suitable for tool crash.
This patch adds file dfi_vmdump.c to handle conversion from
z/VM vmdump format to the internal memory representation of this tool.
Therefore both output formats elf and s390 are supported as target
of the conversion.
The vmdump file formats 32bit and 64bit are not supported anymore.
They have been obsolete since CP version 5.1 (end of service since
30-Sep-2009).
Vector register set is not part of a vmdump file and can not be
displayed.
Output before:
# zgetdump ~/vmdump-0383
zgetdump: No valid dump found on "/root/vmdump-0383"
#
Output after:
# ./zgetdump ~/vmdump-0383 ~/vmdump-0383.elf
Format Info:
Source: vmdump
Target: elf
Copying dump:
00000000 / 00000379 MB
00000379 / 00000379 MB
Success: Dump has been copied
# file ~/vmdump-0383.elf
/root/vmdump-0383.elf: ELF 64-bit MSB core file, IBM S/390, version 1 \
.... (SYSV), SVR4-style, from 'vmlinux'
#
# ./zgetdump -i ~/vmdump-0383
General dump info:
Dump format........: vmdump
Version............: 2
UTS node name......: s8360046.lnxne.boe
UTS kernel release.: 6.1.0-rc2-d-perf+
UTS kernel version.: #1 SMP Fri Oct 28 09:42:13 CEST 2022
System arch........: s390x (64 bit)
CPU count (online).: 2
CPU count (real)...: 2
Dump memory range..: 2048 MB
Memory map:
0000000000000000 - 000000007fffffff (2048 MB)
#
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Sometimes dumping a virtual machine from the outside is the only way to
get the data that is needed. This can be the case if a dumping mechanism
like kdump hasn't been configured or data needs to be fetched at a
specific point. Dumping a protected guest from the outside without help
from FW/HW doesn't yield sufficient data to be useful. Hence we have
introduced Protected Virtualization (PV) dump support - also named
confidential dump support.
The confidential dump support works by integrating the firmware into the
dump process. New Ultravisor calls (UVC) are used to initiate the dump
process, dump CPU data, dump memory state and lastly complete the dump
process. The guest's data is fully encrypted and can only be decrypted
by the entity that owns the customer communication key (CCK) for the
dumped guest. The output format is very similar the normal s390 vmcore
ELF format, it's only enriched by new sections where the returned data
from the UVC "Complete Configuration Dump" and the UVC "Dump
Configuration Storage State" is stored. The encrypted CPU data is stored
in a new note type `NT_S390_PV_CPU_DATA`. The old note types do still
exists but without any confidential data stored. The memory data is
stored in the LOAD segment as usual but for PV dumps it's fully AES-XTS
encrypted.
This commit adds support for reading/decrypting PV guest dumps to
zgetdump by introducing a new DFI input module (`dfi_pv_elf.c`). For
specifying the customer communication key a new command line option
`--key` is added.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This change allows mocking of error/abort macros in unit tests.
Being able to do this in unit tests, enables us to test error conditions w/o
terminating the unit test runner.
The new error functions do not have "noreturn" attribute because
this would make mocking of them in unit tests impossible. We must not
compile these functions as noreturn because we need to return from them
in unit tests and returning from a noreturn function is an undefined
behavior in the C++ standard!
For more details:
- ISO/IEC 14882:2017, Chapter 10.6.8 "Noreturn attribute""
- https://en.cppreference.com/w/cpp/language/attributes/noreturn.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Get rid of C headers which include everything. This is bad for testing and
maintainability. C files should include only what they need and no more.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Create a separate C header for DFI VMCOREINFO to separate
independent things and make it more readable.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Replace magic values with descriptive names to improve readability.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
To prevent segmentation faults when the function dfi_mem_phys_read() is
given invalid parameters, check that the given memory range falls within
a known physical memory chunk, just like it is done for dfi_mem_virt_read().
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The new name describes the purpose of the function better and
emphasizes its difference with the function dfi_mem_phys_read().
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Use util_log from libutil to output various log messages that can be helpful
during problem analysis.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Replace the incorrect %d format specifier with %u one for unsigned integer
types.
This issue was found with AFL fuzzing.
$ ./zdump/zgetdump -i ~/input.bin
General dump info:
Dump format........: s390
Version............: -2147483647
System arch........: s390x (64 bit)
CPU count (online).: 32768
Dump memory range..: 1 MB
Memory map:
0000000000000000 - 00000000000f423f (1 MB)
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Before reading data at addresses 0x10418 and 0x10420 in kdump_init(),
validate the validity of the memory range. Otherwise this can result in
a segmentation fault when zgetdump is given a very small dump, e.g.
S390 DASD single-volume dump of size 0x1000 bytes.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
If a memory chunk is added to mem_phys as well as mem_virt
in dfi_mem_chunk_add_vol() then an illegal memory access might occur
when accessing mem_chunk->data e.g. in dfi_elf_mem_chunk_read_fn()
because the data block pointed to by the data field is now being referenced
by two memory chunks, one in mem_phys and one in mem_virt. If it happens
that the memory chunk from mem_virt is freed in mem_unmap() then
the memory chunk in mem_phys still points to the common data block
which has been already freed. This leads to all sort of bad behavior
in dfi_elf_mem_chunk_read_fn() and other places where mem_chunk->data
might be accessed.
Fixes the following bug:
zgetdump: Unexpected end of file for "dump.all.elf"
And this was found by AddressSanitizer:
=================================================================
==81170==ERROR: AddressSanitizer: heap-use-after-free on address 0x602000000570 at pc 0x00000101ac10 bp 0x03ffd897e250 sp 0x03ffd897e248
READ of size 8 at 0x602000000570 thread T0
#0 0x101ac0f in dfi_elf_mem_chunk_read_fn s390-tools/zdump/dfi_elf.c:27
#1 0x100d8a5 in mem_read s390-tools/zdump/dfi.c:339
#2 0x100d8a5 in dfi_mem_phys_read s390-tools/zdump/dfi.c:616
#3 0x100d8a5 in mem_chunk_map_read_fn s390-tools/zdump/dfi.c:353
#4 0x100fd29 in mem_read s390-tools/zdump/dfi.c:339
#5 0x100fd29 in dfi_mem_read s390-tools/zdump/dfi.c:608
#6 0x1018e89 in os_info_get s390-tools/zdump/dfi_vmcoreinfo.c:65
#7 0x1018e89 in dfi_vmcoreinfo_init s390-tools/zdump/dfi_vmcoreinfo.c:86
#8 0x10175b3 in dfi_init s390-tools/zdump/dfi.c:1215
#9 0x1006e71 in do_stdout s390-tools/zdump/zgetdump.c:161
#10 0x1006e71 in main s390-tools/zdump/zgetdump.c:180
#11 0x3ffb07abb89 in __libc_start_main (/lib64/libc.so.6+0x2bb89)
#12 0x1007e8d (s390-tools/zdump/zgetdump+0x1007e8d)
0x602000000570 is located 0 bytes inside of 8-byte region [0x602000000570,0x602000000578)
freed by thread T0 here:
#0 0x3ffb0bc961b in free (/lib64/libasan.so.6+0xc961b)
#1 0x100d2d9 in mem_unmap s390-tools/zdump/dfi.c:1050
previously allocated by thread T0 here:
#0 0x3ffb0bc9aa9 in calloc (/lib64/libasan.so.6+0xc9aa9)
#1 0x100a271 in zg_alloc s390-tools/zdump/zg.c:93
SUMMARY: AddressSanitizer: heap-use-after-free s390-tools/zdump/dfi_elf.c:27 in dfi_elf_mem_chunk_read_fn
Shadow bytes around the buggy address:
0x100c0400000050: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa 00 fa
0x100c0400000060: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa 00 fa
0x100c0400000070: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa 00 fa
0x100c0400000080: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa 00 fa
0x100c0400000090: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa 00 fa
=>0x100c04000000a0: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa[fd]fa
0x100c04000000b0: fa fa fd fa fa fa fd fa fa fa fd fa fa fa fd fa
0x100c04000000c0: fa fa fd fa fa fa fd fa fa fa fd fa fa fa 04 fa
0x100c04000000d0: fa fa 00 fa fa fa 00 fa fa fa 00 fa fa fa 00 fa
0x100c04000000e0: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x100c04000000f0: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
Shadow byte legend (one shadow byte represents 8 application bytes):
Addressable: 00
Partially addressable: 01 02 03 04 05 06 07
Heap left redzone: fa
Freed heap region: fd
Stack left redzone: f1
Stack mid redzone: f2
Stack right redzone: f3
Stack after return: f5
Stack use after scope: f8
Global redzone: f9
Global init order: f6
Poisoned by user: f7
Container overflow: fc
Array cookie: ac
Intra object redzone: bb
ASan internal: fe
Left alloca redzone: ca
Right alloca redzone: cb
Shadow gap: cc
==81170==ABORTING
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Philipp Rudo <prudo@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add 'Dump file size' field for zgetdump -i output to show the actual size
of dump file in s390 extended format on disk in megabytes.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add verbose option for zgetdump to display the detailed layout of memory
map when printing dump info for s390_ext or ELF dump format.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Update zgetdump tool to process dumps of the new s390 extended format.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
When dump is not complete or dump memory limit is set, zgetdump might end up
with the Segfault on reading Vector Registers. That might take place when
the Vector Registers save area is beyond the dump memory limit.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.vnet.ibm.com>
Reviewed-by: Philipp Rudo <prudo@linux.vnet.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This commit is based on the s390-tools-1.39.0 version.
Changes on top of s390-tools-1.39.0:
- Add MIT license to all source files
- Add LICENSE file
- Transform REAMDE to README.md (markdown)
- Add AUTHORS.md file
- Add CONTRIBUTING.md file
- Move changelog from README to CHANGELOG.md file
Reviewed-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
