linux-kernel - Re: [PATCH 03/22] 2.6.22-rc3 perfmon2 : new system calls support

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

Date:	Thu, 31 May 2007 08:46:15 -0700
From:	Stephane Eranian <eranian@....hp.com>
To:	Christoph Hellwig <hch@...radead.org>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 03/22] 2.6.22-rc3 perfmon2 : new system calls support

Christoph,

On Thu, May 31, 2007 at 04:21:34PM +0100, Christoph Hellwig wrote:
> On Tue, May 29, 2007 at 06:48:17AM -0700, Stephane Eranian wrote:
> > sys_pfm_create_context():
> > 	- create a new perfmon2 context and returns a file descriptor in
> > 	  the pfarg_ctx_t parameters. This is the first call an application
> > 	  must make to do monitoring 
> > 	- rewritten to pass sampling format identification as a string
> > 	- file descriptor is now returned by call
> > 
> > sys_pfm_write_pmcs():
> > 	- program the PMU configuration registers. Accepts vector of arguments
> > 	  of type pfarg_pmc_t
> > 	
> > sys_pfm_write_pmds():
> > 	- program the PMU data registers. Accepts a vector of arguments of type
> > 	  pfarg_pmd_t
> > 
> > sys_pfm_read_pmds():
> > 	- read the PMU data registers.  Accepts a vector of arguments of type
> > 	  pfarg_pmd_t
> 
> This kind of interface doesn't make any sense at all.  Information should
> be read and written from filedescriptors using the read and write family
> syscalls and through the VFS instead of adding tons of system calls.
> 

They are all using file descriptors already.
We use read() for receiving overflow notifications. Write is not used.

> I fear we need to write down the requirements first and then come up
> with something better.  E.g. for per-task sampling an interface centered
> around a few files in /proc/<pid>/ would fit very nicely:
> 
>   /proc/<pid>/perfmon_pmcs
>   /proc/<pid>/perfmon_pmds
>   	Obvious
>   /proc/<pid>/perfmon_ctl
>   	Can get control commands as ascii sets written to
> 
You don't want to do parsing because usually, when sampling, you have
to reprogram the registers on the fly and this is on the critical path.
Information has to be exchanged in binary format.

-- 

-Stephane
-
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/