On 10/29/2018 10:47 PM, Hatayama, Daisuke wrote:
> Currently, in function sadump_kdump_backup_region_init(), variable mem
> holding physical memory to read as a candidate of the ELF core header
> is of type unsigned int with just 4 byte length:
>
> for (i = 0; i < ARRAY_LENGTH(kimage.segment); ++i) {
> char e_ident[EI_NIDENT];
> unsigned mem;
>
> mem=ULONG(buf+i*SIZE(kexec_segment)+OFFSET(kexec_segment.mem));
> if (!mem)
> continue;
>
> if (!readmem(PADDR, mem, e_ident, SELFMAG)) {
> DEBUG_MSG("sadump: failed to read elfcorehdr buffer\n");
> return;
> }
>
> Thus, if backup region for the first 640KB physical memory is located
> at over 4GB location thanks to crashkernel=size,high like:
>
> # grep crashkernel /proc/cmdline
> BOOT_IMAGE=(hd0,gpt2)/vmlinuz-4.18 root=/dev/mapper/rhel-root ro crashkernel=512M,high
>
> # grep Crash /proc/iomem
> 06000000-15ffffff : Crash kernel
> 107f000000-109effffff : Crash kernel
>
> crash> rd -p 0x109ef5d000
> 109ef5d000: 00010102464c457f .ELF....
>
> the upper 32-bit of the physical address in mem variable is dropped
> and readmem() fails while outputting the following debug message:
>
> # LANG=C ./makedumpfile --message-level 8 -f -l -d 31 -x ./vmlinux /dev/sdc vmcore-ld31
> sadump: read dump device as single partition
> sadump: single partition configuration
> page_size : 4096
> sadump: timezone information is missing
> sadump: idtr=fffffe0000000000
> sadump: cr3=86b42e000
> sadump: idtr(phys)=4c35cc000
> sadump: devide_error(vmlinux)=ffffffff81a00c50
> sadump: devide_error(vmcore)=ffffffffa0c00c50
> sadump: cmdline vaddr: ffffffffa1bcf008
> sadump: cmdline paddr: 4c35cf008
> sadump: cmdline buf vaddr: ffff8ae89ffceec0
> sadump: cmdline buf paddr: 109ffceec0
> sadump: kaslr_offset=1f200000
> sadump: phys_base=4a1a00000
> sadump: online cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
> sadump: nr_cpus: 60
>
> sadump: failed to read elfcorehdr buffer <--- This is the debug message indicating
> reading ELF core header fails
>
> Then, the generated vmcore has invalid data in its first 640KB part.
>
> The variable mem needs to have type of 64-bit length.
>
> With this patch, kdump backup region is successfully found as follows:
>
> # LANG=C ./makedumpfile --message-level 31 -f -l -d 31 -x ./vmlinux /dev/sdc vmcore-ld31
> sadump: read dump device as single partition
> sadump: single partition configuration
> page_size : 4096
> sadump: timezone information is missing
> sadump: idtr=fffffe0000000000
> sadump: cr3=86b42e000
> sadump: idtr(phys)=4c35cc000
> sadump: devide_error(vmlinux)=ffffffff81a00c50
> sadump: devide_error(vmcore)=ffffffffa0c00c50
> sadump: cmdline vaddr: ffffffffa1bcf008
> sadump: cmdline paddr: 4c35cf008
> sadump: cmdline buf vaddr: ffff8ae89ffceec0
> sadump: cmdline buf paddr: 109ffceec0
> sadump: kaslr_offset=1f200000
> sadump: phys_base=4a1a00000
> sadump: online cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
> sadump: nr_cpus: 60
> The kernel version is not supported.
> The makedumpfile operation may be incomplete.
> sadump: SRC_START: 0x00000000001000 SRC_SIZE: 0x0000000009f000 SRC_OFFSET: 0x0000109ef61000
> sadump: kdump backup region used
> ...<snip>...
>
> By the way, before crashkernel=size,high was introduced, there was
> limitation that ELF core header resides at under 4GB location, so
> defining it as unsigned int was not entirely wrong at that time.
>
> Signed-off-by: HATAYAMA Daisuke <d.hatayama@xxxxxxxxxxxxxx>
> ---
> sadump_info.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/sadump_info.c b/sadump_info.c
> index dd50d48..a668dca 100644
> --- a/sadump_info.c
> +++ b/sadump_info.c
> @@ -2395,7 +2395,7 @@ sadump_kdump_backup_region_init(void)
> elfcorehdr_p = 0;
> for (i = 0; i < ARRAY_LENGTH(kimage.segment); ++i) {
> char e_ident[EI_NIDENT];
> - unsigned mem;
> + unsigned long mem;
>
> mem=ULONG(buf+i*SIZE(kexec_segment)+OFFSET(kexec_segment.mem));
> if (!mem)
> --
> 2.18.0
>
>
Thanks Hatayama-san, applied to devel branch.
Kazu
_______________________________________________
kexec mailing list
kexec@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/kexec
