From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mga03.intel.com ([134.134.136.65]) by bombadil.infradead.org with esmtps (Exim 4.87 #1 (Red Hat Linux)) id 1cqEFt-0001yZ-6i for linux-mtd@lists.infradead.org; Tue, 21 Mar 2017 07:30:58 +0000 Message-ID: <1490081431.11326.75.camel@gmail.com> Subject: Re: =?UTF-8?Q?=E7=AD=94=E5=A4=8D=3A?= UBIFS: Failed to read/write file when contiguous pages are not enough From: Artem Bityutskiy Reply-To: dedekind1@gmail.com To: =?UTF-8?Q?=E4=BD=99=E5=8D=8E=E5=85=B5=28=E5=B9=B3=E5=8F=B0=E4=B8=80?= =?UTF-8?Q?=E9=83=A8_?= =?UTF-8?Q?=E7=A6=8F=E5=B7=9E=29?= , 'Richard Weinberger' , "linux-mtd@lists.infradead.org" Cc: =?UTF-8?Q?=E4=BF=9E=E5=AF=BF=E5=8D=8E=28=E5=B9=B3=E5=8F=B0=E4=B8=80?= =?UTF-8?Q?=E9=83=A8_?= =?UTF-8?Q?=E7=A6=8F=E5=B7=9E=29?= , =?UTF-8?Q?=E9=82=93=E6=96=87=E6=9D=B0=28=E5=B9=B3=E5=8F=B0=E4=B8=80?= =?UTF-8?Q?=E9=83=A8_?= =?UTF-8?Q?=E7=A6=8F=E5=B7=9E=29?= , =?UTF-8?Q?=E9=BB=84=E5=B4=87=E6=BB=A8=28=E5=B9=B3=E5=8F=B0=E4=B8=80?= =?UTF-8?Q?=E9=83=A8_?= =?UTF-8?Q?=E7=A6=8F=E5=B7=9E=29?= , =?UTF-8?Q?=E6=B1=9F=E4=BF=A1=E8=BE=89=28=E5=B9=B3=E5=8F=B0=E4=B8=80?= =?UTF-8?Q?=E9=83=A8_?= =?UTF-8?Q?=E7=A6=8F=E5=B7=9E=29?= Date: Tue, 21 Mar 2017 09:30:31 +0200 In-Reply-To: <80B78A8B8FEE6145A87579E8435D78C307988C07@FZEX3.ruijie.com.cn> References: <80B78A8B8FEE6145A87579E8435D78C3079814E8@FZEX3.ruijie.com.cn> <1a5eb9d2-8db0-5ca3-d873-a2bd8439c30d@nod.at> <80B78A8B8FEE6145A87579E8435D78C307988C07@FZEX3.ruijie.com.cn> Content-Type: text/plain; charset="UTF-8" Mime-Version: 1.0 Content-Transfer-Encoding: 8bit List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Tue, 2017-03-21 at 01:53 +0000, 余华兵(平台一部 福州) wrote: > Or does this only happen when the kernel tries to produce free memory > in the writeback path? > //This happens when UBIFS reads or writes a file. UBIFS should avoid > applying for continuous pages, otherwise it will fail when the > continuous pages are not enough. You are right that it will be better if UBIFS allocates single pages only. But this is an improvement request. It is not a bug to do order 1 allocations, it is just an imperfection: UBIFS could work better on system with highly fragmented memory. Artem.