Re: [PATCH] UNIX: Do not loop forever at unix_autobind().

["Michał Mirosław" <mirqus@gmail.com>]

2010/9/4 Eric Dumazet <eric.dumazet@gmail.com>:
> Le samedi 04 septembre 2010 à 15:58 +0900, Tetsuo Handa a écrit :
>> From a67ccbb8033993df29f26bde9944e37bffe4fc1b Mon Sep 17 00:00:00 2001
>> From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
>> Date: Sat, 4 Sep 2010 15:22:22 +0900
>> Subject: [PATCH] UNIX: Do not loop forever at unix_autobind().
>>
>> We assumed that unix_autobind() never fails if kzalloc() succeeded.
>> But unix_autobind() allows only 1048576 names. If /proc/sys/fs/file-max is
>> larger than 1048576 (e.g. systems with more than 10GB of RAM), a local user can
>> consume all names using fork()/socket()/bind().
>>
>> If all names are in use, those who call bind() with addr_len == sizeof(short)
>> or connect()/sendmsg() with setsockopt(SO_PASSCRED) will continue
>>
>>   while (1)
>>       yield();
>>
>> loop at unix_autobind() till a name becomes available.
>> This patch changes unix_autobind() to fail if all names are in use.
>>
>> Note that currently a local user can consume 2GB of kernel memory if the user
>> is allowed to create and autobind 1048576 UNIX domain sockets. We should
>> consider adding some restriction for autobind operation.
[cut patch]
> Sorry, this wont work very well if you have many processes using
> autobind(). Some of them will loop many time before hitting
> "stop_ordernum".
>
> unsigned int counter;
>
> ...
>
> if (++maxtries == 1<<20) {
>        ...
> }
>
>
> This is a pathological situation. We are not forced to give a successful
> autobind() when so many sockets are in use, even if some slots are
> available.

Is there any specific requirement on generated names for auto-bind?
Wouldn't it be easier and more efficient to use some pseudo-random
name. i.e. derived from current time and/or owning process state?

Best Regards,
Michał Mirosław