From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from psmtp.com (na3sys010amx147.postini.com [74.125.245.147]) by kanga.kvack.org (Postfix) with SMTP id B7C8E6B0002 for ; Thu, 14 Feb 2013 01:40:52 -0500 (EST) Date: Thu, 14 Feb 2013 01:40:48 -0500 From: Johannes Weiner Subject: Re: [PATCH] mm: Try harder to allocate vmemmap blocks Message-ID: <20130214064048.GB8372@cmpxchg.org> References: <1360816468.5374.285.camel@deadeye.wl.decadent.org.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1360816468.5374.285.camel@deadeye.wl.decadent.org.uk> Sender: owner-linux-mm@kvack.org List-ID: To: Ben Hutchings Cc: linux-mm@kvack.org On Thu, Feb 14, 2013 at 04:34:28AM +0000, Ben Hutchings wrote: > Hot-adding memory on x86_64 normally requires huge page allocation. > When this is done to a VM guest, it's usually because the system is > already tight on memory, so the request tends to fail. Try to avoid > this by adding __GFP_REPEAT to the allocation flags. > > Reported-and-tested-by: Bernhard Schmidt > Reference: http://bugs.debian.org/699913 > Signed-off-by: Ben Hutchings Acked-by: Johannes Weiner > We could go even further and use __GFP_NOFAIL, but I'm not sure > whether that would be a good idea. If __GFP_REPEAT is not enough, I'd rather fall back to regular page backing at this point: diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index 2ead3c8..1f5301d 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c @@ -919,6 +919,7 @@ vmemmap_populate(struct page *start_page, unsigned long size, int node) { unsigned long addr = (unsigned long)start_page; unsigned long end = (unsigned long)(start_page + size); + int use_huge = cpu_has_pse; unsigned long next; pgd_t *pgd; pud_t *pud; @@ -934,8 +935,8 @@ vmemmap_populate(struct page *start_page, unsigned long size, int node) pud = vmemmap_pud_populate(pgd, addr, node); if (!pud) return -ENOMEM; - - if (!cpu_has_pse) { +retry_pmd: + if (!use_huge) { next = (addr + PAGE_SIZE) & PAGE_MASK; pmd = vmemmap_pmd_populate(pud, addr, node); @@ -957,8 +958,10 @@ vmemmap_populate(struct page *start_page, unsigned long size, int node) pte_t entry; p = vmemmap_alloc_block_buf(PMD_SIZE, node); - if (!p) - return -ENOMEM; + if (!p) { + use_huge = 0; + goto retry_pmd; + } entry = pfn_pte(__pa(p) >> PAGE_SHIFT, PAGE_KERNEL_LARGE); -- 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: email@kvack.org