bug-ddrescue
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: reading /proc/self/mem fails


From: Jan Engelhardt
Subject: Re: reading /proc/self/mem fails
Date: Wed, 1 Mar 2023 17:20:48 +0100 (CET)
User-agent: Alpine 2.25 (LSU 592 2021-09-18)

On Wednesday 2023-03-01 16:42, Antonio Diaz Diaz wrote:
>>
>> $ strace -fe lseek ddrescue -f /proc/self/mem /dev/zero -s $[0x4040]
>> ...
>> lseek(3, 0, SEEK_END) = -1 EINVAL (Invalid argument)
>> ...
>>
>> procfs is known to be... special. What is your opinion, is it worth
>> making this work?
>
> Maybe it is worth, but I don't know how to make it work. 'dd' does not work
> either:
> $ dd if=/proc/self/mem of=/dev/zero
> dd: error reading '/proc/self/mem': Input/output error

dd fails to work for a different reason. (Judging from strace output alone
without source review), it uses fstat() instead of lseek(SEEK_END) to determine
the file size -- and then gets confused because stat.st_size < lseek:

$ strace -f dd if=/proc/self/mem of=/dev/zero skip=$[0x4000]
...
lseek(0, 16384, SEEK_CUR) = 16384
newfstatat(0, "", {st_mode=S_IFREG|0600, st_size=0, ...}, AT_EMPTY_PATH) = 0
cannot skip to specified offset

> In fact, nothing seems to work with /proc/self/mem:
> $ cat /proc/self/mem > /dev/zero
> cat: /proc/self/mem: Input/output error

The prerequisite for the special file is that one must seek to a position
where something is mapped. Offset 0, corresponding to mapping address 0x0,
is seldomly mapped.

> $ file /proc/self/mem
> /proc/self/mem: empty

Once again a result of checking st_size.
As said, st_size is 0 because the size of the process cannot be determined.
(Well, it *could* be determined, but perhaps it's seen as "too expensive"
to count all segments whenever the file is stat'ed.)



reply via email to

[Prev in Thread] Current Thread [Next in Thread]