From: Juergen Gross Subject: Re: [PATCH 2/3] xen: correct race in alloc_p2m_pmd() Date: Fri, 09 Jan 2015 16:17:08 +0100 Message-ID: <54AFF0F4.1030608@suse.com> References: <1420736490-15351-1-git-send-email-jgross@suse.com> <1420736490-15351-3-git-send-email-jgross@suse.com> <54AFEF2A.3090403@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <54AFEF2A.3090403@citrix.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: David Vrabel , linux-kernel@vger.kernel.org, xen-devel@lists.xensource.com, konrad.wilk@oracle.com, boris.ostrovsky@oracle.com List-Id: xen-devel@lists.xenproject.org On 01/09/2015 04:09 PM, David Vrabel wrote: > On 08/01/15 17:01, Juergen Gross wrote: >> When allocating a new pmd for the linear mapped p2m list a check is >> done for not introducing another pmd when this just happened on >> another cpu. In this case the old pte pointer was returned which >> points to the p2m_missing or p2m_identity page. The correct value >> would be the pointer to the found new page. > > Looking at the check I don't see how it works at all. > > alloc_p2m() looks up the address of the PTE page > > ptep = lookup_address(addr, &level); > pte_pg = (pte_t *)((unsigned long)ptep & ~(PAGE_SIZE - 1)); > > But the check under the lock that is still true does: > > ptechk = lookup_address(vaddr, &level); > if (ptechk == pte_pg) { > > So I don't see how this works unless (by chance) we happen to get the > first entry in the PTE page. The check under lock tests vaddr which is pmd aligned. > It also doesn't handle racing with p2m_missing_pte to p2m_identity_pte > (or vice-versa) change. The change is always and only for missing/identity to individual pmd. There is no transition between all missing and identity pmds. > Something like: > > ptechk = lookup_address(vaddr, &level); > ptechk = (pte_t *)((unsigned long)ptechk & ~(PAGE_SIZE - 1)); > if (ptechk == p2m_missing_pte || ptechk == p2m_identity) { > set_pmd(..) > > Perhaps? Would work, but is more complicated than needed. :-) Juergen > > David > >> --- a/arch/x86/xen/p2m.c >> +++ b/arch/x86/xen/p2m.c >> @@ -440,10 +440,9 @@ EXPORT_SYMBOL_GPL(get_phys_to_machine); >> * a new pmd is to replace p2m_missing_pte or p2m_identity_pte by a individual >> * pmd. In case of PAE/x86-32 there are multiple pmds to allocate! >> */ >> -static pte_t *alloc_p2m_pmd(unsigned long addr, pte_t *ptep, pte_t *pte_pg) >> +static pte_t *alloc_p2m_pmd(unsigned long addr, pte_t *pte_pg) >> { >> pte_t *ptechk; >> - pte_t *pteret = ptep; >> pte_t *pte_newpg[PMDS_PER_MID_PAGE]; >> pmd_t *pmdp; >> unsigned int level; >> @@ -477,8 +476,6 @@ static pte_t *alloc_p2m_pmd(unsigned long addr, pte_t *ptep, pte_t *pte_pg) >> if (ptechk == pte_pg) { >> set_pmd(pmdp, >> __pmd(__pa(pte_newpg[i]) | _KERNPG_TABLE)); >> - if (vaddr == (addr & ~(PMD_SIZE - 1))) >> - pteret = pte_offset_kernel(pmdp, addr); >> pte_newpg[i] = NULL; >> } >> >> @@ -492,7 +489,7 @@ static pte_t *alloc_p2m_pmd(unsigned long addr, pte_t *ptep, pte_t *pte_pg) >> vaddr += PMD_SIZE; >> } >> >> - return pteret; >> + return lookup_address(addr, &level); >> } >> >> /* >> @@ -521,7 +518,7 @@ static bool alloc_p2m(unsigned long pfn) >> >> if (pte_pg == p2m_missing_pte || pte_pg == p2m_identity_pte) { >> /* PMD level is missing, allocate a new one */ >> - ptep = alloc_p2m_pmd(addr, ptep, pte_pg); >> + ptep = alloc_p2m_pmd(addr, pte_pg); >> if (!ptep) >> return false; >> } >> > >