linux-kernel - msync with MS_ASYNC on same page broken?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <6E48F63F-AB51-4EBE-8A45-E3B5AE8C3BBD@gmail.com>
Date:	Mon, 9 Nov 2009 02:31:17 +0800
From:	Ramneek Handa <ramneekhanda@...il.com>
To:	linux-kernel@...r.kernel.org
Subject: msync with MS_ASYNC on same page broken?

Hi,
I am writing a low latency application which (during the course of  
processing) writes objects into a file in binary mode. I am testing  
the persistence library right now which in single threaded mode can  
achieve about 2 micros per record (if say 100000 records 256 bytes  
each are written and i commit the entire page only once, i.e. one  
commit for 100,000 records).  On the other hand if i write 2 records  
per commit (note there will be 16 records per page), next 15 commits  
will be very slow. If on the other hand i increase my commit size to  
16 per commit the speed becomes 100x.

I read the thread below and looks like the core of the problem.
http://lkml.indiana.edu/hypermail/linux/kernel/0602.1/0500.html

Can some one please shed more light on what decision was eventually  
taken on this and is it already fixed (Andrew's fix seems good for  
such problem)? Any other way i can get around this issue?

BTW, there is no way to predict how many records will be per  
transaction/commit.

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