Re: [PATCH v2][RFC] kdump: append kernel build-id string to VMCOREINFO

[Vijay Balakrishna <vijayb@linux.microsoft.com>]

Thanks Andrew.  See inline.

On 6/17/2020 12:45 PM, Andrew Morton wrote:
> On Wed, 10 Jun 2020 21:27:52 -0700 Vijay Balakrishna <vijayb@linux.microsoft.com> wrote:
> 
>> Make kernel GNU build-id available in VMCOREINFO.  Having
>> build-id in VMCOREINFO facilitates presenting appropriate kernel
>> namelist image with debug information file to kernel crash dump
>> analysis tools.  Currently VMCOREINFO lacks uniquely identifiable
>> key for crash analysis automation.
>>
>> Regarding if this patch is necessary or matching of linux_banner
>> and OSRELEASE in VMCOREINFO employed by crash(8) meets the
>> need -- IMO, build-id approach more foolproof, in most instances it
>> is a cryptographic hash generated using internal code/ELF bits unlike
>> kernel version string upon which linux_banner is based that is
>> external to the code.  I feel each is intended for a different purpose.
>> Also OSRELEASE is not suitable when two different kernel builds
>> from same version with different features enabled.
>>
>> Currently for most linux (and non-linux) systems build-id can be
>> extracted using standard methods for file types such as user mode crash
>> dumps, shared libraries, loadable kernel modules etc.,  This is an
>> exception for linux kernel dump.  Having build-id in VMCOREINFO brings
>> some uniformity for automation tools.
>>
>> ...
>>
>> --- a/kernel/crash_core.c
>> +++ b/kernel/crash_core.c
>> @@ -11,6 +11,8 @@
>>   #include <asm/page.h>
>>   #include <asm/sections.h>
>>   
>> +#include <crypto/sha.h>
>> +
>>   /* vmcoreinfo stuff */
>>   unsigned char *vmcoreinfo_data;
>>   size_t vmcoreinfo_size;
>> @@ -376,6 +378,53 @@ phys_addr_t __weak paddr_vmcoreinfo_note(void)
>>   }
>>   EXPORT_SYMBOL(paddr_vmcoreinfo_note);
>>   
>> +#define NOTES_SIZE (&__stop_notes - &__start_notes)
>> +#define BUILD_ID_MAX SHA1_DIGEST_SIZE
>> +#define NT_GNU_BUILD_ID 3
>> +
>> +struct elf_note_section {
>> +	struct elf_note	n_hdr;
>> +	u8 n_data[];
>> +};
>> +
>> +/*
>> + * Add build ID from .notes section as generated by the GNU ld(1)
>> + * or LLVM lld(1) --build-id option.
>> + */
>> +static void add_build_id_vmcoreinfo(void)
>> +{
>> +	char build_id[BUILD_ID_MAX * 2 + 1];
>> +	int n_remain = NOTES_SIZE;
>> +
>> +	while (n_remain >= sizeof(struct elf_note)) {
>> +		const struct elf_note_section *note_sec =
>> +			&__start_notes + NOTES_SIZE - n_remain;
>> +		const u32 n_namesz = note_sec->n_hdr.n_namesz;
>> +
>> +		if (note_sec->n_hdr.n_type == NT_GNU_BUILD_ID &&
>> +		    n_namesz != 0 &&
>> +		    !strcmp((char *)&note_sec->n_data[0], "GNU")) {
> 
> Is it guaranteed that n_data[] is null-terminated?

 From ELF(5) "Notes (Nhdr)" section:

>        n_namesz  The length of the name field in bytes.  The contents will immediately follow this note in memory.  The name is null  terminated.   For
>                  example, if the name is "GNU", then n_namesz will be set to 4.

Vijay

> 
>> +			if (note_sec->n_hdr.n_descsz <= BUILD_ID_MAX) {
>> +				const u32 n_descsz = note_sec->n_hdr.n_descsz;
>> +				const u8 *s = &note_sec->n_data[n_namesz];
>> +
>> +				s = PTR_ALIGN(s, 4);
>> +				bin2hex(build_id, s, n_descsz);
>> +				build_id[2 * n_descsz] = '\0';
>> +				VMCOREINFO_BUILD_ID(build_id);
>> +				return;
>> +			}
>> +			pr_warn("Build ID is too large to include in vmcoreinfo: %u > %u\n",
>> +				note_sec->n_hdr.n_descsz,
>> +				BUILD_ID_MAX);
>> +			return;
>> +		}
>> +		n_remain -= sizeof(struct elf_note) +
>> +			ALIGN(note_sec->n_hdr.n_namesz, 4) +
>> +			ALIGN(note_sec->n_hdr.n_descsz, 4);
>> +	}
>> +}
>> +
>>   static int __init crash_save_vmcoreinfo_init(void)
>>   {
>>   	vmcoreinfo_data = (unsigned char *)get_zeroed_page(GFP_KERNEL);