From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Howells Subject: Re: [RESEND][PATCH 5/6] Provide a filesystem-specific sync'able page bit Date: Fri, 15 Oct 2004 16:14:46 +0100 Sender: owner-linux-mm@kvack.org Message-ID: <28682.1097853286@redhat.com> References: <20041014194421.GU16153@parcelfarce.linux.theplanet.co.uk> <24461.1097780707@redhat.com> Mime-Version: 1.0 (generated by SEMI 1.14.5 - "Awara-Onsen") Content-Type: text/plain; charset=US-ASCII Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, linux-fsdevel@vger.kernel.org Return-path: In-Reply-To: <20041014194421.GU16153@parcelfarce.linux.theplanet.co.uk> To: Matthew Wilcox List-Id: linux-fsdevel.vger.kernel.org > > +#define PG_fs_misc 9 /* Filesystem specific bit */ > ... > That's not really enough documentation. Who sets this flag? Who clears this > flag? That's up to the owning filesystem (or device file, I suppose); hence "filesystem-specific". I could expand this a little, but there isn't much to say - it's entirely up to the filesystem, though I think I should probably require it to be cleared before the page is freed. > Currently, mm/page_alloc.c clears this flag: Not really; that's irrelevant. It checks to see if it is set when it allocates a page, and if it is it complains bitterly. It then splats this bit and all others to make sure struct page has all its flags in the ground state. > If you're going to not kill this flag, at least rename it so we don't > have two defines for the same bit. I did have it as a different number, but I was told to make it the same as PG_checked. I could rename all instances of PG_checked, I suppose... > which really ought to at least mention Documentation/cachetlb.txt Fix it then. David -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: aart@kvack.org