Re: [PATCH] cifs-utils: smbinfo.c: probably harmless wrong memset sizes + printf format correction

[Pavel Shilovsky <piastryyy@gmail.com>]

Hi Adam,

Sorry it took me a while to look at this. The patch itself looks good
to me. Could you please add an appropriate description, create a patch
with "git format-patch" command and re-send it to the list? This would
allow me to merge it quickly. Submitting a PR on github against the
"next" branch is another good option.

--
Best regards,
Pavel Shilovsky

сб, 25 мая 2019 г. в 15:36, Adam Richter <adamrichter4@gmail.com>:
>
> The attached patch is my attempt at fixing two possibly harmless
> complaints from "cppcheck --enable=warning" from the cifs-utils git
> master branch version of smbinfo.c.
>
> 1. A printf format should have been "%u" instead of "%d" in print_quota.
>
> 2. An incorrect size was being passed to memset in thirteen nearly
> identical places, each using "sizeof(qi)" when "sizeof(*qi)".  I am
> not sure but I think these mistakes were probably harmless because the
> memset calls might all be unnecessary and the sizes passed to each
> memset call might never have been larger than it was supposed to be.
>
> Because each of the effected memset calls was immediately preceded by
> the malloc which allocated the data structure and because they each
> ignored the possibility that malloc could fail, I made a function,
> xmalloc0 to combine allocating the memory, zeroing it and exiting with
> a non-zero exit value and a failure message on allocation failure
> (which appears to be a fine way to handle the error in this program).
> The function uses calloc, which could introduce an unnecessary
> multiply, in the hopes that some calloc implementations may avoid the
> memset in the case of freshly allocated memory from mmap, which would
> probably be the case in this program, although I do not know if any
> calloc implementations make this optimization.  Anyhow, at least this
> way, the size of the data structure is only computed once in the
> source code.
>
> I realize that these memory allocations may all be for small data
> structures that should be allocated on the stack and also may not need
> to be cleared to all zeroes, but I did not want to delve into coding
> style conventions for stack allocation in the CIFS source tree, and I
> was not 100% certain that clearing the allocated memory was
> unnecessary, although I do see other lines that explicitly initialize
> some field in that that allocated memory to zero.  So, please feel
> free to replace my changes with something better or one that involves
> less code churn.
>
> I should also warn that my only testing of these changes was to make
> sure that "cppcheck --enable=warning" no longer complains, that the
> file compiled without complaint (with cifs-utils standard "-Wall
> -Wextra" arguments) and that "./smbinfo quote /dev/null" got past the
> memory allocation to the (correct) ioctl error for /dev/null.
>
> Also, I am not a CIFS developer and this may be the first time I have
> submitted a patch, certainly the first time I remember, so please
> forgive me and feel free to instruct me if I should be following some
> different process to submit this patch.
>
> Thanks in advance for considering this patch submission.
>
> Adam