Re: [Qemu-devel] [PATCH v1] dump: Set correct vaddr for ELF dump

[Laszlo Ersek]

On 01/07/19 13:14, Marc-André Lureau wrote:
> Hi
> 
> On Tue, Dec 25, 2018 at 5:52 PM Jon Doron <address@hidden> wrote:
>>
>> vaddr needs to be equal to the paddr since the dump file represents the
>> physical memory image.
>>
>> Without setting vaddr correctly, GDB would load all the different memory
>> regions on top of each other to vaddr 0, thus making GDB showing the wrong
>> memory data for a given address.
>>
>> Signed-off-by: Jon Doron <address@hidden>
> 
> This is a non-trivial patch! (qemu-trivial, please ignore).
> 
>> ---
>>  dump.c                       | 4 ++--
>>  scripts/dump-guest-memory.py | 1 +
>>  2 files changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/dump.c b/dump.c
>> index 4ec94c5e25..bf77a119ea 100644
>> --- a/dump.c
>> +++ b/dump.c
>> @@ -192,7 +192,7 @@ static void write_elf64_load(DumpState *s, MemoryMapping 
>> *memory_mapping,
>>      phdr.p_paddr = cpu_to_dump64(s, memory_mapping->phys_addr);
>>      phdr.p_filesz = cpu_to_dump64(s, filesz);
>>      phdr.p_memsz = cpu_to_dump64(s, memory_mapping->length);
>> -    phdr.p_vaddr = cpu_to_dump64(s, memory_mapping->virt_addr);
>> +    phdr.p_vaddr = phdr.p_paddr;
> 
> This is likely breaking paging=true somehow, which sets
> memory_mapping->virt_addr to non-0.
> 
> According to doc "If you want to use gdb to process the core, please
> set @paging to true."
> 
> Although I am not able to (gdb) x/10bx 0xa0000 for example on a core
> produced with paging. Not sure why, anybody could help?
> 
>>      assert(memory_mapping->length >= filesz);
>>
>> @@ -216,7 +216,7 @@ static void write_elf32_load(DumpState *s, MemoryMapping 
>> *memory_mapping,
>>      phdr.p_paddr = cpu_to_dump32(s, memory_mapping->phys_addr);
>>      phdr.p_filesz = cpu_to_dump32(s, filesz);
>>      phdr.p_memsz = cpu_to_dump32(s, memory_mapping->length);
>> -    phdr.p_vaddr = cpu_to_dump32(s, memory_mapping->virt_addr);
>> +    phdr.p_vaddr = phdr.p_paddr;
>>
>>      assert(memory_mapping->length >= filesz);
>>
>> diff --git a/scripts/dump-guest-memory.py b/scripts/dump-guest-memory.py
>> index 198cd0fe40..2c587cbefc 100644
>> --- a/scripts/dump-guest-memory.py
>> +++ b/scripts/dump-guest-memory.py
>> @@ -163,6 +163,7 @@ class ELF(object):
>>          phdr = get_arch_phdr(self.endianness, self.elfclass)
>>          phdr.p_type = p_type
>>          phdr.p_paddr = p_paddr
>> +        phdr.p_vaddr = p_paddr
> 
> With your proposed change though, I can dump memory with gdb...
> 
>>          phdr.p_filesz = p_size
>>          phdr.p_memsz = p_size
>>          self.segments.append(phdr)
>> --
>> 2.19.2
>>
>>
> 
> 

I've never used paging-enabled dumps. First, because doing so requires
QEMU to trust guest memory contents (see original commit 783e9b4826b9;
or more recently/generally, the @dump-guest-memory docs in
"qapi/misc.json"). Second, because whenever I had to deal with guest
memory dumps, I always used "crash" (which needs no paging), and the
subject guests were all Linux.

I can't comment on paging-enabled patches for dump, except that they
shouldn't regress the paging-disabled functionality. :) If the patches
satisfy that, I'm fine.

(I *am* surprised that GDB insists on p_vaddr equaling p_paddr; after
all, in the guest, the virtual address is "memory_mapping->virt_addr".
But, I would never claim to understand most of the ELF intricacies,
and/or what GDB requires on top of those.)

Thanks
Laszlo