xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
From: David Vrabel <david.vrabel@citrix.com>
To: "xen-devel@lists.xensource.com" <xen-devel@lists.xensource.com>
Subject: PCI SERR NMI may cause Xen to deadlock
Date: Tue, 31 Jan 2012 16:37:27 +0000	[thread overview]
Message-ID: <4F2818C7.1050702@citrix.com> (raw)

If a PCI System Error (SERR) is asserted it causes an NMI. If this NMI
occurs while the CPU is in printk() then Xen may deadlock as
pci_serr_error() calls console_force_unlock() which screws up the
console lock.

I don't think removing the console_force_unlock() is sufficient as it
looks like printk() is unsafe to call from NMI context.  We would like
keep the diagnostic.  Does the following (untested) patch to defer the
printk() to a softirq look like a valid approach?

I also considered passing the NMI to dom0 (like other NMIs are) but I
couldn't see how NMIs were handled in the current upstream pv-ops kernels.

diff -r e2722b24dc09 xen/arch/x86/traps.c
--- a/xen/arch/x86/traps.c	Thu Jan 26 17:43:31 2012 +0000
+++ b/xen/arch/x86/traps.c	Tue Jan 31 16:28:45 2012 +0000
@@ -3173,6 +3173,11 @@ static void nmi_mce_softirq(void)
     st->vcpu = NULL;
 }

+static void pci_serr_softirq(void)
+{
+    printk("\n\nNMI - PCI system error (SERR)\n");
+}
+
 void async_exception_cleanup(struct vcpu *curr)
 {
     int trap;
@@ -3259,10 +3264,11 @@ static void nmi_dom0_report(unsigned int

 static void pci_serr_error(struct cpu_user_regs *regs)
 {
-    console_force_unlock();
-    printk("\n\nNMI - PCI system error (SERR)\n");
-
     outb((inb(0x61) & 0x0f) | 0x04, 0x61); /* clear-and-disable the PCI
SERR error line. */
+
+    /* Would like to print a diagnostic here but can't call printk()
+       from NMI context -- raise a softirq instead. */
+    raise_softirq(PCI_SERR_SOFTIRQ);
 }

 static void io_check_error(struct cpu_user_regs *regs)
@@ -3563,6 +3569,7 @@ void __init trap_init(void)
     cpu_init();

     open_softirq(NMI_MCE_SOFTIRQ, nmi_mce_softirq);
+    open_softirq(PCI_SERR_SOFTIRQ, pci_serr_softirq);
 }

 long register_guest_nmi_callback(unsigned long address)
diff -r e2722b24dc09 xen/include/asm-x86/softirq.h
--- a/xen/include/asm-x86/softirq.h	Thu Jan 26 17:43:31 2012 +0000
+++ b/xen/include/asm-x86/softirq.h	Tue Jan 31 16:28:45 2012 +0000
@@ -6,6 +6,7 @@
 #define VCPU_KICK_SOFTIRQ      (NR_COMMON_SOFTIRQS + 2)

 #define MACHINE_CHECK_SOFTIRQ  (NR_COMMON_SOFTIRQS + 3)
-#define NR_ARCH_SOFTIRQS       4
+#define PCI_SERR_SOFTIRQ       (NR_COMMON_SOFTIRQS + 4)
+#define NR_ARCH_SOFTIRQS       5

 #endif /* __ASM_SOFTIRQ_H__ */

David

             reply	other threads:[~2012-01-31 16:37 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-01-31 16:37 David Vrabel [this message]
2012-01-31 17:09 ` PCI SERR NMI may cause Xen to deadlock Keir Fraser
2012-02-01 12:03 ` George Dunlap

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=4F2818C7.1050702@citrix.com \
    --to=david.vrabel@citrix.com \
    --cc=xen-devel@lists.xensource.com \
    /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).