Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 30 Sep 2013 07:24:06 -0700
From:      Matthew Fleming <mdf@FreeBSD.org>
To:        Hooman Fazaeli <hoomanfazaeli@gmail.com>
Cc:        FreeBSD Hackers <freebsd-hackers@freebsd.org>
Subject:   Re: rwlock(9) upgrade
Message-ID:  <CAMBSHm8h5LZN49WBwmMhFn-wfs21seZMVf2SgPkWjwdWAYr3sw@mail.gmail.com>
In-Reply-To: <5249158D.5010101@gmail.com>
References:  <5249158D.5010101@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Sep 29, 2013 at 11:09 PM, Hooman Fazaeli <hoomanfazaeli@gmail.com>wrote:

> Hi,
>
> For rwlock(9), there is no rwlock_upgrade function.
> Is it safe to use rw_wlock() for that purpose? In other words, Does calling
> rw_wlock() upgradeanalready r-locked lock?



No, calling rw_wlock when you hold the lock in read mode will deadlock.

lockmgr(9) supports LK_UPGRADE, but note (see the recent commit for
LK_TRYUPGRADE) that LK_UPGRADE, if it cannot do the upgrade immediately,
e.g. due to multiple shared lockers, will unlock and wait to finish the
upgrade.

In general, the idea is that upgrade is not a good operation, since there's
no way to know ahead of time if it can be done without a lock release.  So
code is better off explicitly unlocking the shared/read-mode lock and
explicitly blocking for an exclusive/write lock.

Thanks,
matthew



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAMBSHm8h5LZN49WBwmMhFn-wfs21seZMVf2SgPkWjwdWAYr3sw>