From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.codeaurora.org ([198.145.11.231]) by bombadil.infradead.org with esmtps (Exim 4.80.1 #2 (Red Hat Linux)) id 1WeNwe-00055C-FO for linux-mtd@lists.infradead.org; Sun, 27 Apr 2014 12:12:32 +0000 Message-ID: <789ed0034a7f4aa993addb2bc8c7e914.squirrel@www.codeaurora.org> In-Reply-To: <1396871016.17040.33.camel@sauron.fi.intel.com> References: <1396260879.9016.70.camel@sauron.fi.intel.com> <533A5407.4060900@codeaurora.org> <1396871016.17040.33.camel@sauron.fi.intel.com> Date: Sun, 27 Apr 2014 12:12:11 -0000 Subject: Re: ubifs: assertion fails From: "Dolev Raviv" To: dedekind1@gmail.com MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit Cc: Dolev Raviv , linux-mtd@lists.infradead.org, Tanya Brokhman List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , > On Tue, 2014-04-01 at 08:52 +0300, Tanya Brokhman wrote: >> The log isn't flooded with the above mentioned error, but it does repeat >> several times. > > Even though you do a lot of I/O operations? > > If the accounting becomes incorrect, I think you'd see a warning for > each and every I/O operations. Yes, the log seems clean to me. The error actually occurs while stressing with iozone benchmark. Could the error be a result of the corner case you described in your previous mail. > > Also, do you have several UBIFS file-systems mounted? > Yes, we have prepared images based on ubifs. So almost all the mounted partitions are ubifs. >> Is there any way we can help in debugging and fixing this? Also, we're >> running on a 3.10 based kernel and I saw a lot of patches that change >> the shrinker after 3.10 on linux-next. What kernel version did you see >> the shrinker errors on? > > Well, you just need to stress the shrinker. I did it by simply adding a > hack to ubifs module init which kmalloc()'ed a lot of memory. Not > too-much, but enough to cause some pressure when doing a lot of I/O > operations and get UBIFS shrinker be invoked. This took me some amount > of time to get a good hack. I use a normal PC with some Fedora, and > nandsime. This makes it a lot easier to debug, comparing with doing this > on a real target. After all, shrinker is a generic component. I think > that PC had 4GB of RAM. I do not remember how much I allocated in my > hack. I'll probably will take a look at this once I finish other more urgent tasks. > > I think I used older version than 3.10 back then. > > -- > Best Regards, > Artem Bityutskiy > > -- QUALCOMM ISRAEL, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation