| Age | Commit message (Collapse) | Author |
|---|
|
This reverts commit 5edcbfd1368e84fce913ceeeca7b712c524dc20d.
Yinghai Lu has reported on the kexec mailing list that this causes
the following problem when using kexec load with kexec built on
openSUSE 13.1 64bit.
overflow in relocation type R_X86_64_32 val 21dffc020
|
|
Currently, the two options must be the last ones to be honored;
otherwise, they can get silently ignored and both the manpage and help
text point it out. This is error-prone and trivial to fix. There
isn't much point in pointing something out in documentation when the
peculiarity can be removed with four lines of extra code.
Update option handling so that the two arguments are honored
regardless of their positions.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
During kernel develoment, kexec can be used to get out of sticky
situations without going through possibly lenghty reboot; however,
there are situations where the filesystem and/or storage stack are
known to be misbehaving and performing sync before kexecing is
dangerous or just never finishes.
This patch implement -y (--no-sync) option which makes kexec skip
syncing in the similar way as -x (--no-ifdown).
Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Chris Mason <clm@fb.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Summary of changes,
configure.ac: Add test for detect x32 ABI.
purgatory/arch/x86_64/Makefile: Not use mcmodel large when
x32 ABI is set.
kexec/arch/x86_64/kexec-elf-rel-x86_64.c: When x32 ABI is set
use ELFCLASS32 instead of ELFCLASS64.
kexec/kexec-syscall.h: Add correct syscall number for x32 ABI.
Signed-off-by: Aníbal Limón <anibal.limon@linux.intel.com>
Signed-off-by: Mariano Lopez <mariano.lopez@linux.intel.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
gcc 4.9.1 tells me this variable is set but unused
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Commit 4362bfac changes params for kexec_iomem_for_each_line from
'unsigned long' to 'unsigned long long'.
This patch fixes forgotten changes for sh and x86 archs.
Bug causes incorrect parsing of memory ranges.
Signed-off-by: Roman Pen <r.peniaev@gmail.com>
Cc: kexec@lists.infradead.org
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
kexec/arch/i386/kexec-elf-x86.c:97:6: warning: variable
‘modified_cmdline_len’ set but not used [-Wunused-but-set-variable]
int modified_cmdline_len;
Signed-off-by: Ameya Palande <2ameya@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
kexec/arch/i386/kexec-bzImage.c:111:8: warning: variable
‘kernel_version’ set but not used [-Wunused-but-set-variable]
Signed-off-by: Ameya Palande <2ameya@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
The sizeof operator returns a size_t type. Change the printf format of sizeof
values from %d to %zu. Fixes compiler warnings like these:
kexec/fs2dt.c: warning: format ‘%d’ expects argument of type ‘int’, but argument 2 has type ‘long unsigned int’ [-Wformat=]
Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Add braces to if statments in fs2dt.c. Fixes warnings like these when compiling
with gcc 4.9.1:
kexec/fs2dt.c: warning: suggest explicit braces to avoid ambiguous ‘else’ [-Wparentheses]
Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Fixes the following compilation warning.
kexec/arch/mips/crashdump-mips.c:151:6: warning: unused variable 'i' [-Wunused-variable]
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Fixes the following warning.
kexec/arch/mips/kexec-elf-mips.c:161:16: warning: assignment makes integer from pointer without a cast [enabled by default]
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
If "--command-line" option is not specified, then kexec segfaults while
dereferencing NULL command line string pointer. While we are at it, also
fix indentation and use '{' and '}' consistently.
Signed-off-by: Ameya Palande <2ameya@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Let {zlib,lzma}_decompress_file() return NULL if anything wrong happened
to allow the other method to have a chance to run.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
In ppc64 loading, it will print the segment information. This is not
wanted since other Archs don't have this. People even think there's
something wrong. So erase it to make it be consistent with other Archs.
And if people really want to check these info, they can specify "-d"
option. They are printed in print_segments() too.
Signed-off-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
This reverts commit 930f0c3734c223433e74037601cbfd1bc9df7bad.
|
|
panic_on_warn kernel parameter will cause the kernel to panic when a
WARN() is hit in the kernel. This is not a good situation for the kdump
kernel because then it would be possible for the kdump kernel to panic in
a non-fatal WARN().
This patch removes panic_on_warn as a kernel parameter for the kdump
kernel.
Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: WANG Chao <chaowang@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
The OPAL firmware is going to embed its symbol map in the device tree.
The size is large enough to be more than a page, and it takes
multiple reads to get the whole file. This is because sysfs uses
the seq_file helpers which do a page at a time.
Unfortunately fs2dt has no handling for short reads and we die with:
unrecoverable error: short read from"/proc/device-tree//ibm,opal/firmware/symbol-map"
This patch uses the slurp_file_len helper which does the right thing.
It moves the explicit open of the file further down for
add_usable_mem_property and add_dyn_reconf_usable_mem_property.
We should convert both of these to use the buffer provided by
slurp_file_len at some stage.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
since we have already cleared kexec_info with memset,
there is no need to do that again to the struct members.
Signed-off-by: Hu Keping <hukeping@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Commit b02d735bf252 ('ppc64 kdump device_tree sort') added code to
sort device tree properties, but it had a few issues.
A compare routine needs to return -1 and 1. The special case for
sorting properties with unit addresses only returned 1 and ignored
the opposite case, which screwed up the sorting.
We were missing a few more things:
- Need to check both basenames are the same length
- Need to check both basenames match
I noticed this when looking at the NUMA topology after a kexec, and
it had shifted.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Currently kexec returns success even if an invalid command line option
is encountered. Kexec currently prints usage message and then returns
0. That's not right. It is an error and error code 1 should be returned.
Due to this wrapper script thinks that kdump succeeded but that's not
the case.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
The ELF header created among the loading of the kdump kernel should be
flagged using the current endianess and not always as big endian.
Without this patch the data exposed in /proc/vmcore are not readable when
running in LE mode.
Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Hi,
This is v2 of the patch. Since v1, I moved syscall implemented check littler
earlier in the function as per the feedback.
Now a new kexec syscall (kexec_file_load()) has been merged in upstream
kernel. This system call takes file descriptors of kernel and initramfs
as input (as opposed to list of segments to be loaded). This new system
call allows for signature verification of the kernel being loaded.
One use of signature verification of kernel is secureboot systems where
we want to allow kexec into a kernel only if it is validly signed by
a key system trusts.
This patch provides and option --kexec-file-syscall (-s), to force use of
new system call for kexec. Default is to continue to use old syscall.
Currently only bzImage64 on x86_64 can be loaded using this system call.
As kernel adds support for more arches and for more image types, kexec-tools
can be modified accordingly.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Acked-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
When loading dtb on arm platform, we may need change strings in dtb
in function setup_dtb_prop. If length of new string is larger than that
of the old one. We should extend the dtb by "FDT_TAGALIGN"ed size.
If not, kexec may fail to load the capture kernel.
Signed-off-by: Liu Hua <sdu.liu@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Kernel commit e6a6928c3ea1d0195ed75a091e345696b916c09b changed the way the
device tree is processed in the kernel. Now version 2 is no more supported.
This patch move the version of the device tree generated in ppc64
environment from 2 to 17, allowing to kexec kernel 3.16.
In addition, automates the define of NEED_STRUCTURE_BLOCK_EXTRA_PAD which
should not be set for DT version 16 and above.
Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
For 32-bit ARM systems with CONFIG_ARM_LPAE=y, when kexec utility
loads the crash kernel. 32-bit elf header is not enough if the
physical address exceeds 4G.
This patch checkes whether the largest physical address of the system
exceeds 4G. If so, kexec creates 64-bit elf header.Otherwise it
creates 32-bit elf header.
Signed-off-by: Wei Jitao <weijitao@huawei.com>
Signed-off-by: Liu Hua <sdu.liu@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Since "start" and "end" in kexec_iomem_for_each_line
are unsigned long long, this patch makes "callback"s
accept unsigned-long-long-type parameters.
Signed-off-by: Wei Jitao <weijitao@huawei.com>
Signed-off-by: Liu Hua <sdu.liu@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
get_crash_notes_per_cpu() should return -1 if return value of stat(2) is
zero (on success).
Signed-off-by: Takao Indoh <indou.takao@jp.fujitsu.com>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: WANG Chao <chaowang@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
There is a case that the number of /sys/devices/system/cpu/cpuN is not
contiguous, for example after cpu hot removing. This patch fixes so that
all /sys/devices/system/cpu/cpuN is handled when they are discontiguous.
Signed-off-by: Takao Indoh <indou.takao@jp.fujitsu.com>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: WANG Chao <chaowang@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
If kernel does not export efi runtime maps it means 1:1 mapping does not
work or user explictly boot with efi=old_map. In this case efi setup code
will failback to noefi boot, but for kdump case we still need pass extra
acpi_rsdp cmdline.
Thus adding a check in kdump path.
Signed-off-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
When KALLSYMS is disabled or /proc is not mounted, this patch allows
--page-offset cmdline option for user to setup correct page_offset.
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
ARM kernel can be compiled with CONFIG_VMSPLIT_1G, CONFIG_VMSPLIT_2G or
CONFIG_VMSPLIT_3G. This patch dynamically detects PAGE_OFFSET according
to _stext symbol from /proc/kallsyms.
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
There's one more '-' in arch_usage, thus
s/pass--memmap-cmdline/pass-memmap-cmdline
Signed-off-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
For kernel boot with efi=old_map or some quirked machines like SGI UV
they use old ioremap instead of 1:1 mapping. But kexec efi support
depends on the 1:1 mapping thus we need to switch to use the old way
There's a kernel patch for exporting the efi flags so we can check the memory
mapping method. But user may want to explictly disable efi boot for unknown
reasons.
So here add a new arch option '--noefi' for this case.
Signed-off-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
In kdump path, now we store all the 2nd kernel memory ranges in
memmap_p. We could use just cmdline_add_memmap() to add all types of
memory ranges to 2nd kernel cmdline.
So clean up here, merge cmdline_add_memmap_acpi() into
cmdline_add_memmap().
Signed-off-by: WANG Chao <chaowang@redhat.com>
Acked-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
OPAL based system exports reserved memory ranges through /proc/device-tree
for the regions that are reserved by OPAL firmware. Traverse
/proc/device-tree/reserved-ranges and add them to exclude_ranges[] and
reserve them.
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
On PowerNV platform, OPAL region is overlapped with crashkernel, need to
create ELF Program header for the overlapped memory. The changes are
similar to the way RTAS region was handled.
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
This patch memset modified_cmdline after it alloced.
Without this patch, the header of command line always contains random
value like following example:
kernel: 0x765fd008 kernel_size: 0x23b610
memory range: [0x60000000 - 0x6fffffff] (256M)
memory range: [0x78000000 - 0x9fffffff] (640M)
...
kernel command line: "°¥氥䐶# elfcorehdr=0x77f00000 mem=130048K"
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Geng Hui <hui.geng@huawei.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
I accidentally add one duplicate line. Now remove it.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
This patch fixs a problem introduced by commit e5d6a55 which make ARM
kexec fails.
Due to that commit, kernel is loaded at a dynamically offset: it computes
extra_size using size of dtb, and load zImage at base + extra_size. When
dtb size small (for example, 0x3000 bytes), kernel will be loaded at
address like 0x60003000. For ARM zImage such address is incorrect.
In kernel code arch/arm/boot/compressed/head.S, zImage builds a
temporary page table at (pc & 0xf8000000) + TEXT_OFFSET - 0x4000. The
related instructions sequence is:
mov r4, pc
and r4, r4, #0xf8000000
add r4, r4, #TEXT_OFFSET @ (TEXT_OFFSET == 0x8000 on most platforms)
call cache_on
...
call __armv7_mmu_cache_on
...
call __setup_mmu
sub r3, r4, #16384 @ Page directory size
r3 becomes page table pointer.
When kernel is loaded at 0x60003000, page table is still built at
0x60004000, which destroys kernel.
This patch make extra_size a fix value (0x8000) to avoid the failure.
For the problem commit e5d6a55 tries solve, this patch uses
locate_holes() to find a place can hold initrd and dtb.
Change from v1:
- Coding style enhancements.
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
In commit 91f5b9c ("kdump: pass e820 reserved region to 2nd kernel via
e820 table or setup_data"), I made a wrong condition check.
We should only add cmdline for a memory range if --pass-memmap-cmdline
and the range type isn't RANGE_RESERVED.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
The code in arch/arm/kexec-zImage-arm.c currently enforces a hard limit
on the maximum size a dtb blob can occupy. This limit is set to 32k,
which is quite low for device tree blobs nowadays.
Get rid of this assumption, and calculate the added size dynamically.
For this, we need to slurp in the dtb file earlier in order to
determine its size, because the memory hole allocation for 'base'
takes this size into account.
For ATAGs, we keep the current value of 32k, which should in fact be
enough.
With this change in place, the 'DTB too large!' error message can go
away. Successfully tested on a AM335x board.
Signed-off-by: Daniel Mack <zonque@gmail.com>
[s.neumann@raumfeld.com: Fix ATAGs case]
Reported-and-tested-by: Sven Neumann <s.neumann@raumfeld.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
e820 reserved region could be useful in 2nd kernel.
For example, PCI mmconf (extended mode) requires reserved region
otherwise it falls back to legacy mode. The following log is from Cliff
Wickman <cpw@sgi.com>:
PCI: MMCONFIG for domain 1003 [bus 3f-3f] at [mem 0xff0ff00000-0xff0fffffff] (base 0xff0c000000)
[Firmware Bug]: PCI: MMCONFIG at [mem 0x80000000-0x80cfffff] not reserved in ACPI motherboard resources
PCI: not using MMCONFIG
PCI devices on segment 1 (>0) can't fall back to legacy mode, thus
kernel probing fails and device can't be found.
We don't pass reserved region because these regions could be too much
and eat up our very limited kernel command line resource in
memmap=exactmap case.
However now we use e820 map and setup_data to pass memory map to 2nd
kernel and the number of reserved regions should not be a problem any
more.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Generated files should always be built from source and never be present in
VCS repositories and only autotools generated files should be in tarballs.
This ensures that they get built as often as possible and bugs with that
process are discovered early.
Signed-off-by: Paul Wise <pabs3@bonedaddy.net>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
command line size is restricted by kernel, sometimes memmap=exactmap has
too many memory ranges to pass to cmdline. And also memmap=exactmap and
kASLR doesn't work together.
A better approach, to pass the memory ranges for crash kernel to boot
into, is filling the memory ranges into E820.
boot_params only got 128 slots for E820 map to fit in, when the number of
memory map exceeds 128, use setup_data to pass the rest as extended E820
memory map.
kexec boot could also benefit from setup_data in case E820 memory map
exceeds 128.
Now this new approach becomes default instead of memmap=exactmap.
saved_max_pfn users can specify --pass-memmap-cmdline to use the
exactmap approach.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Tested-by: Linn Crosetto <linn@hp.com>
Reviewed-by: Linn Crosetto <linn@hp.com>
Acked-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
--pass-memmap-cmdline is used for pass memmap=exactmap cmdline for 2nd
kernel. Later we will use this option to disable passing E820 memmap
method but use the old exactmap method.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Tested-by: Linn Crosetto <linn@hp.com>
Acked-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Later kexec and kdump memory range will be mapped to E820entry. But
currently kexec memory range .end field is exclusive while crash memory
range is inclusive.
Given the fact that the exported proc iomem and sysfs memmap are both
inclusive, change kexec memory range .end to be inclusive. Later the
unified memory range of both kexec and kdump can use the same E820
filling code.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Acked-by: Dave Young <dyoung@redhat.com>
Tested-by: Linn Crosetto <linn@hp.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
|
Add two new members to kexec_info structure:
struct memory_range *crash_range
int nr_crash_ranges;
crash_range contains the memory ranges used to boot 2nd kernel.
nr_crash_ranges contains the count of the crash memory ranges.
Signed-off-by: WANG Chao <chaowang@redhat.com>
Acked-by: Dave Young <dyoung@redhat.com>
Tested-by: Linn Crosetto <linn@hp.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
|
