public inbox for [email protected]  

help / color / mirror / Atom feed

From: Tomas Vondra <[email protected]>
To: Heikki Linnakangas <[email protected]>
To: Ranier Vilela <[email protected]>
To: Nathan Bossart <[email protected]>
To: Tomas Vondra <[email protected]>
Cc: Pg Hackers <[email protected]>
Subject: Re: Avoid possible overflow (src/port/bsearch_arg.c)
Date: Mon, 28 Oct 2024 16:30:26 +0100
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>
References: <CAEudQAp34o_8u6sGSVraLwuMv9F7T9hyHpePXHmRaxR2Aboi+w@mail.gmail.com>
	<ZwWkBb2gxs5fGttK@nathan>
	<CAEudQAr8KZET1ogPnT7esb7YdYbmJyFZWo-OnP+FgoKeL+N31A@mail.gmail.com>
	<[email protected]>



On 10/28/24 13:13, Heikki Linnakangas wrote:
> On 09/10/2024 19:16, Ranier Vilela wrote:
>> Em ter., 8 de out. de 2024 às 18:28, Nathan Bossart
>> <[email protected] <mailto:[email protected]>> escreveu:
>>
>>     On Tue, Oct 08, 2024 at 04:09:00PM -0300, Ranier Vilela wrote:
>>      > The port function *bsearch_arg* mimics the C function
>>      > *bsearch*.
>>      >
>>      > The API signature is:
>>      > void *
>>      > bsearch_arg(const void *key, const void *base0,
>>      > size_t nmemb, size_t size,
>>      > int (*compar) (const void *, const void *, void *),
>>      > void *arg)
>>      >
>>      > So, the parameter *nmemb* is size_t.
>>      > Therefore, a call with nmemb greater than INT_MAX is possible.
>>      >
>>      > Internally the code uses the *int* type to iterate through the
>>     number of
>>      > members, which makes overflow possible.
>>
>>     I traced this back to commit bfa2cee (v14), which both moved
>>     bsearch_arg()
>>     to its current location and adjusted the style a bit.  Your patch
>> looks
>>     reasonable to me.
>>
>> Thanks for looking.
> 
> Committed, thanks.
> 
> Based on the original discussion on bfa2cee, I couldn't figure out where
> exactly this new bsearch implementation originated from, but googling
> around, probably *BSD or libiberty. Tomas, do you remember? Not that it
> matters, but I'm curious.
> 

I don't remember, unfortunately :-( I think it was one of the *BSDs,
because of license, but I'm not quite sure why I changed the code at all
during the move.

> Some of those other implementations have fixed this, others have not.
> And they all seem to also have the "involes" typo in the comment that we
> fixed in commit 7ef8b52cf07 :-). Ranier, you might want to submit this
> fix to those other projects too.
> 

Thanks for fixing this, although I wonder if we can actually hit this,
as we don't really allocate more than 1GB in most places. But it's
possible, and the pre-bfa2cee code handled it fine.


regards

-- 
Tomas Vondra

view thread (10+ messages)  latest in thread

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: [email protected]
  Cc: [email protected], [email protected], [email protected], [email protected], [email protected]
  Subject: Re: Avoid possible overflow (src/port/bsearch_arg.c)
  In-Reply-To: <[email protected]>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox