Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 29 May 2006 13:52:48 +0900
From:      gnn@FreeBSD.org
To:        Robert Watson <rwatson@FreeBSD.org>, freebsd-atm@FreeBSD.org, freebsd-arch@FreeBSD.org
Subject:   Re: Locking netatm
Message-ID:  <m23betzb33.wl%gnn@neville-neil.com>
In-Reply-To: <20060528230058.GA836@lucy.pool-70-17-33-65.pskn.east.verizon.net>
References:  <20060528230058.GA836@lucy.pool-70-17-33-65.pskn.east.verizon.net>

next in thread | previous in thread | raw e-mail | index | archive | help
At Sun, 28 May 2006 19:00:59 -0400,
Skip Ford wrote:
> 
> In my attempt to make netatm MPSAFE, I've run into just a couple
> situations I'm not sure how to handle.  I've more-or-less replaced
> the splnet() calls with a single subsystem lock.  Nearly all of them
> were protecting structures within the netatm subsystem.  But I'm not
> sure how to handle the existing splimp() calls.
> 
Hi Skip,

I'm not familiar with netatm but I am somewhat familiar with our
locking code, and I hope that others will also join in.

> There seem to be a few different reasons why splimp() calls were
> made.  First, for timing.  Three netatm functions use splimp() to
> protect access to the list of atm_timeq structures (struct atm_time
> *).  Here is the usage from one of those three functions.  I've
> removed much of the code to simplify what it does:
> 
> 	s = splimp();
> 
> 	FOREACH(atm_timeq);
> 		...
> 
> 	... possibly modify structures within atm_timeq ...
> 	... modify the struct atm_time * passed to this function ...
> 
> 	(void) splx(s);
> 
> So my question is, were network interrupts disabled when mucking
> with the atm_timeq list because a generated interrupt can modify
> structures within the list?  This use is probably very
> netatm-specific.  I'm still studying the timeout code to understand
> what it's doing.
> 

The spl() calls haven't disabled real interrupts, as far as I know,
for quite a while.  They acted as general code locks to prevent
simultaneous access to data structures while an update was in
progress.  In terms of the timeq, the locks were acting as a mutex now
would to protect the list during an update.

> A second situation where network interrupts were disabled was for
> netatm memory allocation for devices:
> 
> in atm_dev_alloc()
> 
> 	s = splimp();
> 
> 	FOREACH(atm_mem_head)
> 		...
> 	malloc (...)
> 
> 	(void) splx(s);
> 
> and in atm_dev_free()
> 
> 	s = splimp();
> 
> 	FOREACH(atm_mem_head)
> 		...
> 	free (...);
> 
> 	(void) splx(s);
> 
> I'm not sure how these should be protected.  Presumably, we don't
> want to receive interrupts until the netatm memory for the device is
> allocated.  Would a global subsystem lock protect these calls?  I
> can protect atm_mem_head, so maybe that'd be enough?

I would assume, again, that these are prtections of the list, and not
the memory allocation routines.  A mutex protecting the list, like the
one that protects, say, the UDP protocol list is probably what you want.

> Another use is to protect calls to other subsystems.  For
> example:
> 
> within atm_nif_attach(struct atm_nif *nip)
> 
> 	ifp = nip->nif_ifp;
> 
> 	s = splimp();
> 
> 	if_attach(ifp);
> 	bpfattach(ifp, DLT_ATM_CLIP, T_ATM_LLC_MAX_LEN);
> 
> 	(void) splx(s);
> }
> 
> and within atm_nif_detach(struct atm_nif *nip)
> 
> 	ifp = nip->nif_ifp;
> 
> 	s = splimp();
> 
> 	bpfdetach(ifp);
> 	if_detach(ifp);
> 	if_free(ifp);
> 
> 	(void) splx(s);
> 
> Holding a new netatm subsystem lock won't protect those calls so I'm
> not sure how to handle those.  Other non-netatm code in the tree
> seems to not do any locking at all around those calls.

I believe these are now unnecessary since the ifnet lists now have
their own locks.

> These are really the only uses I've yet to convert so if someone can
> provide some pointers, I'd appreciate it.  I'm pretty new to FreeBSD
> locking, either the old way or the new way.  I'm still studying the
> code, including other network stacks and the netatm stack itself,
> but a pointer or two would be appreciated.  I feel like it's mostly
> converted, though I've done no testing at all yet.  Once I finish
> removing splimp(), I can test with the single subsystem lock, then
> move on to finer-grained locking where necessary.

I would look at the UDP and TCP protocol list locking as being
somewhat similar to what you have here.  UDP is the easiest to
understand as it's mostly in one file, netinet/udp_usrreq.c.

Later,
George




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?m23betzb33.wl%gnn>