From owner-svn-src-head@freebsd.org Thu Jun 14 19:44:13 2018 Return-Path: Delivered-To: svn-src-head@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 782931020D0B; Thu, 14 Jun 2018 19:44:13 +0000 (UTC) (envelope-from kib@freebsd.org) Received: from kib.kiev.ua (kib.kiev.ua [IPv6:2001:470:d5e7:1::1]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 022137DAEA; Thu, 14 Jun 2018 19:44:12 +0000 (UTC) (envelope-from kib@freebsd.org) Received: from tom.home (kib@localhost [127.0.0.1]) by kib.kiev.ua (8.15.2/8.15.2) with ESMTP id w5EJi2FU078420; Thu, 14 Jun 2018 22:44:05 +0300 (EEST) (envelope-from kib@freebsd.org) DKIM-Filter: OpenDKIM Filter v2.10.3 kib.kiev.ua w5EJi2FU078420 Received: (from kostik@localhost) by tom.home (8.15.2/8.15.2/Submit) id w5EJi2VB078419; Thu, 14 Jun 2018 22:44:02 +0300 (EEST) (envelope-from kib@freebsd.org) X-Authentication-Warning: tom.home: kostik set sender to kib@freebsd.org using -f Date: Thu, 14 Jun 2018 22:44:02 +0300 From: Konstantin Belousov To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r335171 - head/sys/vm Message-ID: <20180614194402.GJ2493@kib.kiev.ua> References: <201806141941.w5EJf2qa069373@repo.freebsd.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <201806141941.w5EJf2qa069373@repo.freebsd.org> User-Agent: Mutt/1.10.0 (2018-05-17) X-Spam-Status: No, score=-2.9 required=5.0 tests=ALL_TRUSTED,BAYES_00 autolearn=ham autolearn_force=no version=3.4.1 X-Spam-Checker-Version: SpamAssassin 3.4.1 (2015-04-28) on tom.home X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.26 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 14 Jun 2018 19:44:13 -0000 On Thu, Jun 14, 2018 at 07:41:02PM +0000, Konstantin Belousov wrote: > Author: kib > Date: Thu Jun 14 19:41:02 2018 > New Revision: 335171 > URL: https://svnweb.freebsd.org/changeset/base/335171 > > Log: > Handle the race between fork/vm_object_split() and faults. > > If fault started before vmspace_fork() locked the map, and then during > fork, vm_map_copy_entry()->vm_object_split() is executed, it is > possible that the fault instantiate the page into the original object > when the page was already copied into the new object (see > vm_map_split() for the orig/new objects terminology). This can happen > if split found a busy page (e.g. from the fault) and slept dropping > the objects lock, which allows the swap pager to instantiate > read-behind pages for the fault. Then the restart of the scan can see > a page in the scanned range, where it was already copied to the upper > object. > > Fix it by instantiating the read-ahead pages before > swap_pager_getpages() method drops the lock to allocate pbuf. The > object scan would see the whole range prefilled with the busy pages > and not proceed the range. > > Note that vm_fault rechecks the map generation count after the object > unlock, so that it restarts the handling if raced with split, and > re-lookups the right page from the upper object. > > In collaboration with: alc Reviewed by: markj Sorry. > Tested by: pho > Sponsored by: The FreeBSD Foundation > MFC after: 1 week Differential revision: https://reviews.freebsd.org/D15293 > > Modified: > head/sys/vm/swap_pager.c > > Modified: head/sys/vm/swap_pager.c > ============================================================================== > --- head/sys/vm/swap_pager.c Thu Jun 14 19:01:40 2018 (r335170) > +++ head/sys/vm/swap_pager.c Thu Jun 14 19:41:02 2018 (r335171) > @@ -1096,21 +1096,24 @@ swap_pager_getpages(vm_object_t object, vm_page_t *ma, > int *rahead) > { > struct buf *bp; > - vm_page_t mpred, msucc, p; > + vm_page_t bm, mpred, msucc, p; > vm_pindex_t pindex; > daddr_t blk; > - int i, j, maxahead, maxbehind, reqcount, shift; > + int i, maxahead, maxbehind, reqcount; > > reqcount = count; > > - VM_OBJECT_WUNLOCK(object); > - bp = getpbuf(&nsw_rcount); > - VM_OBJECT_WLOCK(object); > - > - if (!swap_pager_haspage(object, ma[0]->pindex, &maxbehind, &maxahead)) { > - relpbuf(bp, &nsw_rcount); > + /* > + * Determine the final number of read-behind pages and > + * allocate them BEFORE releasing the object lock. Otherwise, > + * there can be a problematic race with vm_object_split(). > + * Specifically, vm_object_split() might first transfer pages > + * that precede ma[0] in the current object to a new object, > + * and then this function incorrectly recreates those pages as > + * read-behind pages in the current object. > + */ > + if (!swap_pager_haspage(object, ma[0]->pindex, &maxbehind, &maxahead)) > return (VM_PAGER_FAIL); > - } > > /* > * Clip the readahead and readbehind ranges to exclude resident pages. > @@ -1132,35 +1135,31 @@ swap_pager_getpages(vm_object_t object, vm_page_t *ma, > *rbehind = pindex - mpred->pindex - 1; > } > > + bm = ma[0]; > + for (i = 0; i < count; i++) > + ma[i]->oflags |= VPO_SWAPINPROG; > + > /* > * Allocate readahead and readbehind pages. > */ > - shift = rbehind != NULL ? *rbehind : 0; > - if (shift != 0) { > - for (i = 1; i <= shift; i++) { > + if (rbehind != NULL) { > + for (i = 1; i <= *rbehind; i++) { > p = vm_page_alloc(object, ma[0]->pindex - i, > VM_ALLOC_NORMAL); > - if (p == NULL) { > - /* Shift allocated pages to the left. */ > - for (j = 0; j < i - 1; j++) > - bp->b_pages[j] = > - bp->b_pages[j + shift - i + 1]; > + if (p == NULL) > break; > - } > - bp->b_pages[shift - i] = p; > + p->oflags |= VPO_SWAPINPROG; > + bm = p; > } > - shift = i - 1; > - *rbehind = shift; > + *rbehind = i - 1; > } > - for (i = 0; i < reqcount; i++) > - bp->b_pages[i + shift] = ma[i]; > if (rahead != NULL) { > for (i = 0; i < *rahead; i++) { > p = vm_page_alloc(object, > ma[reqcount - 1]->pindex + i + 1, VM_ALLOC_NORMAL); > if (p == NULL) > break; > - bp->b_pages[shift + reqcount + i] = p; > + p->oflags |= VPO_SWAPINPROG; > } > *rahead = i; > } > @@ -1171,15 +1170,18 @@ swap_pager_getpages(vm_object_t object, vm_page_t *ma, > > vm_object_pip_add(object, count); > > - for (i = 0; i < count; i++) > - bp->b_pages[i]->oflags |= VPO_SWAPINPROG; > - > - pindex = bp->b_pages[0]->pindex; > + pindex = bm->pindex; > blk = swp_pager_meta_ctl(object, pindex, 0); > KASSERT(blk != SWAPBLK_NONE, > ("no swap blocking containing %p(%jx)", object, (uintmax_t)pindex)); > > VM_OBJECT_WUNLOCK(object); > + bp = getpbuf(&nsw_rcount); > + /* Pages cannot leave the object while busy. */ > + for (i = 0, p = bm; i < count; i++, p = TAILQ_NEXT(p, listq)) { > + MPASS(p->pindex == bm->pindex + i); > + bp->b_pages[i] = p; > + } > > bp->b_flags |= B_PAGING; > bp->b_iocmd = BIO_READ;