From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ed L Cashin Date: Tue, 16 Mar 2004 21:03:23 +0000 Subject: Re: REAL_PTRS_PER_PMD and TIF_32BIT Message-Id: <87fzc88pyc.fsf@uga.edu> List-Id: References: <87llm08qho.fsf@uga.edu> In-Reply-To: <87llm08qho.fsf@uga.edu> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: sparclinux@vger.kernel.org "David S. Miller" writes: > On Tue, 16 Mar 2004 15:51:47 -0500 > Ed L Cashin wrote: > >> Hi. As far as I can tell, PTRS_PER_PMD will evaluate to 512 when a >> 32-bit binary in userland makes a system call or has a page fault, >> since TIF_32BIT will be set in the thread flags. >> >> I suppose that means that for every second-level page table associated >> with a 32-bit binary's process, there are (2048 - 512) unused >> pmd entries. Is that correct? > > Yes, on sparc64, this is what effectively happens. We only need one > PMD to map the page tables for a 32-bit process on sparc64. I notice that in the kernel "pmd" is used to mean both the second level page table, consisting of 2048 four-byte entries, and also as shorthand for one of those entries. Did you intentionally use all-caps for "PMD" to distinguish it from a second-level page table entry? -- --Ed L Cashin | PGP public key: ecashin@uga.edu | http://noserose.net/e/pgp/