Re: [PATCH 1/4] module: add syscall to load module from fd

[Lucas De Marchi]

On Sun, Jan 6, 2013 at 4:59 PM, Michael Kerrisk (man-pages)
<mtk.manpages@xxxxxxxxx> wrote:
> Hi Rusty, (and Lucas, and Kees)
>
> On Thu, Jan 3, 2013 at 1:12 AM, Rusty Russell <rusty@xxxxxxxxxxxxxxx> wrote:
>> Michael Kerrisk <mtk.manpages@xxxxxxxxx> writes:
>>> Hi Rusty,
>>
>> Hi Michael,
>>
>>> The description here is rather thin. Could you supply a sentence or
>>> two for each of MODULE_INIT_IGNORE_MODVERSIONS and
>>> MODULE_INIT_IGNORE_VERMAGIC that would be suitable for the manual
>>> page?
>>>
>>> Thanks,
>>
>> There are one or two safety checks built into a module, which are
>> checked to match the kernel on module load.  The first is a "vermagic"
>> string containing the kernel version number and prominent features (such
>> as CPU type).  If the module was built with CONFIG_MODVERSIONS set, a
>> version hash is recorded for each symbol the module uses based on the
>> types it refers to: in this case, the kernel version number within the
>> "vermagic" string is ignored, as the symbol version hashes are assumed
>> to be sufficiently reliable.
>>
>> Using the MODULE_INIT_IGNORE_VERMAGIC flag indicates that the vermagic
>> is to be ignored, and the MODULE_INIT_IGNORE_MODVERSIONS flag indicates
>> that the version hashes are to be ignored.  If the kernel is built to
>> permit such forced loading (ie. CONFIG_MODULE_FORCE_LOAD is set) then
>> loading will continue, otherwise it will fail with ENOEXEC as expected
>> for malformed modules.
>>
>> Hope that is more usable?
>
> Yes, that helps. I did some reworking of that text. Hopefully, I did
> not introduce any errors.
>
> Below is the text that is proposed to document finit_module() in the
> man pages. I'd appreciate any review (Kees, Lucas, Rusty?)
>
> Thanks,
>
> Michael
>
>    finit_module()
>        The finit_module() system call is like init_module(), but reads
>        the module to be loaded from the file  descriptor  fd.   It  is
>        useful  when  the authenticity of a kernel module can be deterâ
>        mined from its location in the file system; in cases where that
>        is  possible,  the  overhead  of using cryptographically signed
>        modules to determine  the  authenticity  of  a  module  can  be
>        avoided.  The param_values argument is as for init_module().
>
>        The  flags  argument  modifies the operation of finit_module().
>        It is a bit mask value created by ORing together zero  or  more
>        of the following flags:
>
>        MODULE_INIT_IGNORE_MODVERSIONS
>               Ignore symbol version hashes.
>
>        MODULE_INIT_IGNORE_VERMAGIC
>               Ignore kernel version magic.
>
>        There are some safety checks built into a module to ensure that
>        it matches the kernel against which it is loaded.  These checks
>        are  recorded  when  the  module is built and verified when the
>        module is loaded.   First,  the  module  records  a  "vermagic"
>        string  containing the kernel version number and prominent feaâ
>        tures (such as the CPU type).  Second, if the module was  built
>        with  the  CONFIG_MODVERSIONS  configuration  option enabled, a
>        version hash is recorded for each symbol the module uses.  This
>        hash  is  based  on the types of the arguments and return value
>        for the function named by the symbol.  In this case, the kernel
>        version  number within the "vermagic" string is ignored, as the
>        symbol version hashes are assumed to be sufficiently reliable.
>
>        Using the MODULE_INIT_IGNORE_VERMAGIC flag indicates  that  the
>        "vermagic"   string   is   to   be   ignored,   and   the  MODâ
>        ULE_INIT_IGNORE_MODVERSIONS flag indicates that the symbol verâ
>        sion  hashes are to be ignored.  If the kernel is built to perâ
>        mit  forced  loading   (i.e.,   configured   with   CONFIG_MODâ
>        ULE_FORCE_LOAD),  then loading will continue, otherwise it will
>        fail with ENOEXEC as expected for malformed modules.
> ...
>    ERRORS
> ...
>        The following errors may additionally occur for finit_module():
>
>        EBADF  The file referred to by fd is not opened for reading.
>
>        EFBIG  The file referred to by fd is too large.
>
>        EINVAL flags is invalid.
>
>        ENOEXEC
>               fd does not refer to an open file.
>
>


Looks good to me.


regards,
Lucas De Marchi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/