From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:48298) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bP3PG-0003u7-4I for qemu-devel@nongnu.org; Mon, 18 Jul 2016 03:56:03 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1bP3PD-0006GE-29 for qemu-devel@nongnu.org; Mon, 18 Jul 2016 03:56:02 -0400 Received: from mx-v6.kamp.de ([2a02:248:0:51::16]:54807 helo=mx01.kamp.de) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bP3PC-0006G1-OA for qemu-devel@nongnu.org; Mon, 18 Jul 2016 03:55:58 -0400 References: <1467284842-22458-1-git-send-email-pl@kamp.de> <08e4d9ef-8d72-e4fd-5494-7a9301bf6438@redhat.com> From: Peter Lieven Message-ID: <578C8B81.9020509@kamp.de> Date: Mon, 18 Jul 2016 09:55:45 +0200 MIME-Version: 1.0 In-Reply-To: <08e4d9ef-8d72-e4fd-5494-7a9301bf6438@redhat.com> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH V4] block/iscsi: allow caching of the allocation map List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Paolo Bonzini , qemu-block@nongnu.org Cc: kwolf@redhat.com, famz@redhat.com, qemu-devel@nongnu.org, mreitz@redhat.com, ronniesahlberg@gmail.com Am 16.07.2016 um 15:38 schrieb Paolo Bonzini: > > On 30/06/2016 13:07, Peter Lieven wrote: >> +static void >> +iscsi_allocmap_update(IscsiLun *iscsilun, int64_t sector_num, >> + int nb_sectors, bool allocated, bool valid) >> { >> int64_t cluster_num, nb_clusters; >> - if (iscsilun->allocationmap == NULL) { >> + >> + if (iscsilun->allocmap == NULL) { >> return; >> } >> cluster_num = DIV_ROUND_UP(sector_num, iscsilun->cluster_sectors); >> nb_clusters = (sector_num + nb_sectors) / iscsilun->cluster_sectors >> - cluster_num; >> - if (nb_clusters > 0) { >> - bitmap_clear(iscsilun->allocationmap, cluster_num, nb_clusters); >> + if (allocated) { >> + bitmap_set(iscsilun->allocmap, >> + sector_num / iscsilun->cluster_sectors, >> + DIV_ROUND_UP(nb_sectors, iscsilun->cluster_sectors)); >> + } else if (nb_clusters > 0) { >> + bitmap_clear(iscsilun->allocmap, cluster_num, nb_clusters); > I'm sorry for the delay in review, but this is still wrong. > > Suppose cluster_sectors is 2, sector_num = 1, nb_sectors = 6: > > 0--.--2--.--4--.--6--.--8 > |_________________| > > Here in the "mark unallocated" case you want to set 2..6, i.e. > cluster_num=2, nb_clusters=2. This is right. > > 0--.--2--.--4--.--6--.--8 > xxx|___________|xxx (x = shrunk) cluster_num = 1 and nb_clusters = 2 > > In the "mark allocated" case, instead, you want to set 0..8, i.e. > cluster_num=0, nb_clusters=4. > > 0--.--2--.--4--.--6--.--8 > <--|_________________|--> (<--> = expanded) > > Instead you are setting nb_clusters=3, so that 6..8 is not marked > allocated and reading 6..7 will return zeroes: > > 0--.--2--.--4--.--6--.--8 > <--|______________|!!! (! = wrong) > > Instead you need to use > > DIV_ROUND_UP(sector_num + nb_sectors, ...) > - sector_num / iscsilun->cluster_sectors > > which does the rounding correctly. Right, but this has been wrong even in the currently active version. So I would propose to send 2 patches. 1) fix the rounding in the old version cc'ing qemu-stable 2) rebase the new patch on top of that. Thanks for catching this, Peter