Date: Tue, 12 Oct 2004 15:30:06 -0700 (PDT) From: Don Lewis <truckman@FreeBSD.org> To: alfred@FreeBSD.org Cc: jhb@FreeBSD.org Subject: Re: cvs commit: src/sys/kern kern_subr.c Message-ID: <200410122230.i9CMU6dv073297@gw.catspoiler.org> In-Reply-To: <20041012221950.GT38364@elvis.mu.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 12 Oct, Alfred Perlstein wrote: > * Don Lewis <truckman@FreeBSD.org> [041012 15:17] wrote: >> On 12 Oct, Alfred Perlstein wrote: >> > I like this, my only concern is that there may be places that >> > call this with locks held but with the kernel/user buffer wired >> > so that it can't fault. >> > >> > Are you sure this isn't the case? (specifically for some sysctls) >> >> There are a number of sysctl handlers where I added code to wire the >> buffer so that there was no danger of sleeping while a mutex is held. >> Allocating a kernel buffer and doing an extra copy would be ugly. > > Yup. Maybe an auxilliary uio_move_wired() that doesn't do this > check? As I mentioned in my followup message, the sysctl handlers don't use uiomove(). This might be a good idea for copyin() and copyout(), where the _wired versions would check to see if the memory was actually wired.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200410122230.i9CMU6dv073297>