Debian Bug report logs - #441387
Misleading mprotect manpage

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: François Diakhate <diakhate@enseirb.fr>

To: submit@bugs.debian.org

Subject: Misleading mprotect manpage

Date: Sun, 09 Sep 2007 13:37:06 +0200

Package: manpages-dev
Version: 2.39-1

The mprotect manpage specifies that:

"Whether PROT_EXEC has any effect different from PROT_READ is  
architecture and kernel version dependent."

which is true, but it should also specify that whether PROT_WRITE 
implies PROT_READ is architecture dependant and that it is the case on 
x86 platforms. The same holds true for the mmap manpage. However, what 
makes this manpage especially misleading in that regard is that it 
states that:

"For  example,  if the memory had previously been marked PROT_READ, and 
mprotect() is then called with prot PROT_WRITE, it will no longer be 
readable."

which, as I said, is not true on x86 platforms. I think would be better 
to reverse this sentence so as to say:

"For  example,  if the memory had previously been marked PROT_WRITE, and 
mprotect() is then called with prot PROT_READ, it will no longer be 
writable."

Message #10 received at 441387@bugs.debian.org (full text, mbox, reply):

From: Michael Kerrisk <mtk-manpages@gmx.net>

To: François Diakhate <diakhate@enseirb.fr>, 441387@bugs.debian.org

Cc: debc <control@bugs.debian.org>

Subject: Re: Bug#441387: Misleading mprotect manpage

Date: Mon, 10 Sep 2007 06:10:48 +0200

tags 441387 fixed-upstream
thanks

Hello François,

> The mprotect manpage specifies that:
> 
> "Whether PROT_EXEC has any effect different from PROT_READ is 
> architecture and kernel version dependent."
> 
> which is true, but it should also specify that whether PROT_WRITE
> implies PROT_READ is architecture dependant and that it is the case on
> x86 platforms.

Yes, agreed.  I added a sentence:

    Whether PROT_EXEC has any effect different from  PROT_READ  is
    architecture  and  kernel version dependent.  On some hardware
    architectures (e.g., x86), PROT_WRITE implies PROT_READ.

> The same holds true for the mmap manpage.

I also added a similar sentence to mmap(2).

> However, what
> makes this manpage especially misleading in that regard is that it
> states that:
> 
> "For  example,  if the memory had previously been marked PROT_READ, and
> mprotect() is then called with prot PROT_WRITE, it will no longer be
> readable."
> 
> which, as I said, is not true on x86 platforms. I think would be better
> to reverse this sentence so as to say:
> 
> "For  example,  if the memory had previously been marked PROT_WRITE, and
> mprotect() is then called with prot PROT_READ, it will no longer be
> writable."

I cannot find this text in the upstream man page (is it in mmap.2
or mprotect.2?).  Is this some text added by Debian?

Cheers,

Michael

-- 
Michael Kerrisk
maintainer of Linux man pages Sections 2, 3, 4, 5, and 7

Want to help with man page maintenance?  Grab the latest tarball at
http://www.kernel.org/pub/linux/docs/manpages/
read the HOWTOHELP file and grep the source files for 'FIXME'.

Message #17 received at 441387@bugs.debian.org (full text, mbox, reply):

From: François Diakhate <diakhate@enseirb.fr>

To: 441387@bugs.debian.org

Subject: Re: Bug#441387: Misleading mprotect manpage

Date: Mon, 10 Sep 2007 14:12:25 +0200

Michael Kerrisk wrote:
> tags 441387 fixed-upstream
> thanks
>
> Hello François,
>
>   
Hello Michael,
>> The mprotect manpage specifies that:
>>
>> "Whether PROT_EXEC has any effect different from PROT_READ is 
>> architecture and kernel version dependent."
>>
>> which is true, but it should also specify that whether PROT_WRITE
>> implies PROT_READ is architecture dependant and that it is the case on
>> x86 platforms.
>>     
>
> Yes, agreed.  I added a sentence:
>
>     Whether PROT_EXEC has any effect different from  PROT_READ  is
>     architecture  and  kernel version dependent.  On some hardware
>     architectures (e.g., x86), PROT_WRITE implies PROT_READ.
>
>   
>> The same holds true for the mmap manpage.
>>     
>
> I also added a similar sentence to mmap(2).
>   

Great !
>   
>> However, what
>> makes this manpage especially misleading in that regard is that it
>> states that:
>>
>> "For  example,  if the memory had previously been marked PROT_READ, and
>> mprotect() is then called with prot PROT_WRITE, it will no longer be
>> readable."
>>
>> which, as I said, is not true on x86 platforms. I think would be better
>> to reverse this sentence so as to say:
>>
>> "For  example,  if the memory had previously been marked PROT_WRITE, and
>> mprotect() is then called with prot PROT_READ, it will no longer be
>> writable."
>>     
>
> I cannot find this text in the upstream man page (is it in mmap.2
> or mprotect.2?).  Is this some text added by Debian?
>   
Actually, the text can be found in the upstream mprotect.2  man page 
version 2.39 but it looks like it has been removed from the latest versions.

Thanks,
François
> Cheers,
>
> Michael
>
>   

Send a report that this bug log contains spam.