[dm-devel] [PATCH v3 02/19] libmultipath: fix tur checker double locking

[Benjamin Marzinski]

On Fri, Oct 05, 2018 at 12:25:15PM +0200, Martin Wilck wrote:
> On Thu, 2018-10-04 at 11:45 -0500,  Benjamin Marzinski  wrote:
> > On Mon, Oct 01, 2018 at 10:09:41PM +0200, Martin Wilck wrote:
> > > On Fri, 2018-09-21 at 18:05 -0500, Benjamin Marzinski wrote:
> > > > tur_devt() locks ct->lock. However, it is ocassionally called
> > > > while
> > > > ct->lock is already locked. In reality, there is no reason why we
> > > > need
> > > > to lock all the accesses to ct->devt. The tur checker only needs
> > > > to
> > > > write to this variable one time, when it first gets the file
> > > > descripter
> > > > that it is checking.  It also never uses ct->devt directly.
> > > > Instead,
> > > > it
> > > > always graps the major and minor, and turns them into a string.
> > > > This
> > > > patch changes ct->devt into that string, and sets it in
> > > > libcheck_init()
> > > > when it is first initializing the checker context. After that,
> > > > ct-
> > > > > devt
> > > > 
> > > > is only ever read.
> > > 
> > > I like the lock removal a lot, but not so much the conversion into
> > > a
> > > string. Why not keep the dev_t? 
> > 
> > Because we will simply convert it into a string every time we use it,
> > instead of doing the work one time. It's 24 more bytes in the
> > tur_checker_context, but the code is easier to read, and we're not
> > doing
> > the same work again and again.
> 
> Well, IMO snprintf(buf, N, "%d:%d", major(x), minor(x)) is not _that_
> much work, and it looks a lot cleaner, to me at least.
> 
> But OK, I'm not insisting on this. So, if you re-post, I'm going to ack
> the patch.
> 
> My thought for long term is is this actually this: ct->dev_t is only
> needed for creating messages to be stored in the checker->message
> string, which I don't like either.
> 
> Why don't we replace the "message" field with an "int msgid", and
> add a 
> 
>      char* (*get_message)(int msgid)
> 
> to the checker API?

I would not have a problem with a patch that did this.

-Ben

> 
> msgid could actually be another "uatomic" field. We could even go one
> step further, create a field for the checker status in struct checker,
> and use like one byte for the general path status and the rest of the
> field for the msgid, which usually represents just some checker-
> specific extra information wrt the status.
> 
> > > 
> > > Or maybe even easier, the other way around: why don't we make it a
> > > char* and simply set checker->dev_t = &pp->dev_t?
> > 
> > The whole reason we have tur_checker_context->holders is that it's
> > possible for a path to be removed (or orphaned) while the thread is
> > still running. The tur_checker_context needs to keep all its own
> > storage, so that it never as to worry about pointing to freed memory.
> 
> Yeah, I got that meanwhile, as posted before. Sorry.
> 
> Martin
> 
> -- 
> Dr. Martin Wilck <mwilck at suse.com>, Tel. +49 (0)911 74053 2107
> SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton
> HRB 21284 (AG Nürnberg)
>