From: Tamas K Lengyel <tamas.lengyel@intel.com>
To: xen-devel@lists.xenproject.org
Cc: "Tamas K Lengyel" <tamas.lengyel@intel.com>,
"Wei Liu" <wl@xen.org>,
"Andrew Cooper" <andrew.cooper3@citrix.com>,
"George Dunlap" <george.dunlap@citrix.com>,
"Jan Beulich" <jbeulich@suse.com>,
"Roger Pau Monné" <roger.pau@citrix.com>
Subject: [PATCH for-4.14] x86/hap: use get_gfn_type in hap_update_paging_modes
Date: Tue, 16 Jun 2020 12:31:06 -0700 [thread overview]
Message-ID: <6a2ae3bae4a4ad32bc7caecd8af2655a76a9fb19.1592335579.git.tamas.lengyel@intel.com> (raw)
While forking VMs running a small RTOS systems (Zephyr) a Xen crash has been
observed due to a mm-lock order violation while copying the HVM CPU context
from the parent. This issue has been identified to be due to
hap_update_paging_modes getting a lock on the gfn using get_gfn. This call also
creates a shared entry in the fork's memory map for the cr3 gfn. The function
later calls hap_update_cr3 while holding the paging_lock, which results in the
lock-order violation in vmx_load_pdptrs when it tries to unshare the above entry.
This issue has not affected VMs running other OS's as a call to vmx_load_pdptrs
is benign if PAE is not enabled or if EFER_LMA is set and returns before
trying to unshare and map the page.
Using get_gfn_type to get a lock on the gfn avoids this problem as we can
populate the fork's gfn with a copied page instead of a shared entry if its
needed, thus avoiding the lock order violation while holding paging_lock.
Signed-off-by: Tamas K Lengyel <tamas.lengyel@intel.com>
---
The bug seems to have been present since commit 4cb6c4f4941, only discovered
now due to the heavy use of mem_sharing with VM forks. As this is a simple
bug-fix it would be nice to include it in the 4.14 release.
---
xen/arch/x86/mm/hap/hap.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)
diff --git a/xen/arch/x86/mm/hap/hap.c b/xen/arch/x86/mm/hap/hap.c
index 7f84d0c6ea..9ae4c3ae6e 100644
--- a/xen/arch/x86/mm/hap/hap.c
+++ b/xen/arch/x86/mm/hap/hap.c
@@ -748,12 +748,19 @@ static void hap_update_paging_modes(struct vcpu *v)
struct domain *d = v->domain;
unsigned long cr3_gfn = v->arch.hvm.guest_cr[3] >> PAGE_SHIFT;
p2m_type_t t;
+ p2m_query_t q = P2M_ALLOC;
- /* We hold onto the cr3 as it may be modified later, and
- * we need to respect lock ordering. No need for
- * checks here as they are performed by vmx_load_pdptrs
- * (the potential user of the cr3) */
- (void)get_gfn(d, cr3_gfn, &t);
+ /*
+ * We hold onto the cr3 as it may be modified later, and
+ * we need to respect lock ordering. Unshare here if we have to as to avoid
+ * a lock-order violation later while we are holding the paging_lock.
+ * Further checks are performed by vmx_load_pdptrs (the potential user of
+ * the cr3).
+ */
+ if ( hvm_pae_enabled(v) && !hvm_long_mode_active(v) )
+ q |= P2M_UNSHARE;
+
+ (void)get_gfn_type(d, cr3_gfn, &t, q);
paging_lock(d);
v->arch.paging.mode = hap_paging_get_mode(v);
--
2.25.1
next reply other threads:[~2020-06-16 19:31 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-06-16 19:31 Tamas K Lengyel [this message]
2020-06-17 8:23 ` [PATCH for-4.14] x86/hap: use get_gfn_type in hap_update_paging_modes Roger Pau Monné
2020-06-17 12:49 ` Tamas K Lengyel
2020-06-17 14:29 ` Roger Pau Monné
2020-06-17 14:45 ` Tamas K Lengyel
2020-06-17 9:58 ` Jan Beulich
2020-06-17 13:00 ` Tamas K Lengyel
2020-06-17 13:04 ` Jan Beulich
2020-06-17 13:21 ` Tamas K Lengyel
2020-06-17 13:28 ` Jan Beulich
2020-06-17 13:31 ` Tamas K Lengyel
2020-06-17 13:36 ` Jan Beulich
2020-06-17 13:43 ` Tamas K Lengyel
2020-06-17 14:24 ` Jan Beulich
2020-06-17 14:49 ` Tamas K Lengyel
2020-06-17 15:46 ` Jan Beulich
2020-06-17 15:54 ` Tamas K Lengyel
2020-06-17 15:59 ` Tamas K Lengyel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=6a2ae3bae4a4ad32bc7caecd8af2655a76a9fb19.1592335579.git.tamas.lengyel@intel.com \
--to=tamas.lengyel@intel.com \
--cc=andrew.cooper3@citrix.com \
--cc=george.dunlap@citrix.com \
--cc=jbeulich@suse.com \
--cc=roger.pau@citrix.com \
--cc=wl@xen.org \
--cc=xen-devel@lists.xenproject.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).