[Libvir] PATCH: Fix compat for Xen 3.2.0
Xen 3.2.0 removed the sockets_per_node field from the nodeinfo data returned by XenD. I previously add compat for this, but got it in the wrong place so it would most likely end up in a divide-by-zero error. This patch re-arranges it to be correct. diff -rup libvirt-0.4.0.orig/src/xend_internal.c libvirt-0.4.0.new/src/xend_internal.c --- libvirt-0.4.0.orig/src/xend_internal.c 2007-12-17 18:05:27.000000000 -0500 +++ libvirt-0.4.0.new/src/xend_internal.c 2008-01-18 21:13:30.000000000 -0500 @@ -1907,6 +1907,9 @@ sexpr_to_xend_node_info(struct sexpr *ro info->mhz = sexpr_int(root, "node/cpu_mhz"); info->nodes = sexpr_int(root, "node/nr_nodes"); info->sockets = sexpr_int(root, "node/sockets_per_node"); + info->cores = sexpr_int(root, "node/cores_per_socket"); + info->threads = sexpr_int(root, "node/threads_per_core"); + /* Xen 3.2.0 replaces sockets_per_node with 'nr_cpus'. * Old Xen calculated sockets_per_node using its internal * nr_cpus / (nodes*cores*threads), so fake it ourselves @@ -1921,8 +1924,6 @@ sexpr_to_xend_node_info(struct sexpr *ro if (info->sockets == 0) info->sockets = 1; } - info->cores = sexpr_int(root, "node/cores_per_socket"); - info->threads = sexpr_int(root, "node/threads_per_core"); return (0); } Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
"Daniel P. Berrange" <berrange@redhat.com> wrote:
Xen 3.2.0 removed the sockets_per_node field from the nodeinfo data returned by XenD. I previously add compat for this, but got it in the wrong place so it would most likely end up in a divide-by-zero error. This patch re-arranges it to be correct.
diff -rup libvirt-0.4.0.orig/src/xend_internal.c libvirt-0.4.0.new/src/xend_internal.c --- libvirt-0.4.0.orig/src/xend_internal.c 2007-12-17 18:05:27.000000000 -0500 +++ libvirt-0.4.0.new/src/xend_internal.c 2008-01-18 21:13:30.000000000 -0500 @@ -1907,6 +1907,9 @@ sexpr_to_xend_node_info(struct sexpr *ro info->mhz = sexpr_int(root, "node/cpu_mhz"); info->nodes = sexpr_int(root, "node/nr_nodes"); info->sockets = sexpr_int(root, "node/sockets_per_node"); + info->cores = sexpr_int(root, "node/cores_per_socket"); + info->threads = sexpr_int(root, "node/threads_per_core"); + /* Xen 3.2.0 replaces sockets_per_node with 'nr_cpus'. * Old Xen calculated sockets_per_node using its internal * nr_cpus / (nodes*cores*threads), so fake it ourselves @@ -1921,8 +1924,6 @@ sexpr_to_xend_node_info(struct sexpr *ro if (info->sockets == 0) info->sockets = 1; } - info->cores = sexpr_int(root, "node/cores_per_socket"); - info->threads = sexpr_int(root, "node/threads_per_core"); return (0); }
ACK. Now that you mention divide-by-zero, is there something that guarantees none of the terms in the denominator there is zero? info->sockets = nr_cpus / (info->nodes * info->cores * info->threads);
On Sat, Jan 19, 2008 at 09:21:07PM +0100, Jim Meyering wrote:
"Daniel P. Berrange" <berrange@redhat.com> wrote:
Xen 3.2.0 removed the sockets_per_node field from the nodeinfo data returned by XenD. I previously add compat for this, but got it in the wrong place so it would most likely end up in a divide-by-zero error. This patch re-arranges it to be correct.
diff -rup libvirt-0.4.0.orig/src/xend_internal.c libvirt-0.4.0.new/src/xend_internal.c --- libvirt-0.4.0.orig/src/xend_internal.c 2007-12-17 18:05:27.000000000 -0500 +++ libvirt-0.4.0.new/src/xend_internal.c 2008-01-18 21:13:30.000000000 -0500 @@ -1907,6 +1907,9 @@ sexpr_to_xend_node_info(struct sexpr *ro info->mhz = sexpr_int(root, "node/cpu_mhz"); info->nodes = sexpr_int(root, "node/nr_nodes"); info->sockets = sexpr_int(root, "node/sockets_per_node"); + info->cores = sexpr_int(root, "node/cores_per_socket"); + info->threads = sexpr_int(root, "node/threads_per_core"); + /* Xen 3.2.0 replaces sockets_per_node with 'nr_cpus'. * Old Xen calculated sockets_per_node using its internal * nr_cpus / (nodes*cores*threads), so fake it ourselves @@ -1921,8 +1924,6 @@ sexpr_to_xend_node_info(struct sexpr *ro if (info->sockets == 0) info->sockets = 1; } - info->cores = sexpr_int(root, "node/cores_per_socket"); - info->threads = sexpr_int(root, "node/threads_per_core"); return (0); }
ACK.
Now that you mention divide-by-zero, is there something that guarantees none of the terms in the denominator there is zero?
info->sockets = nr_cpus / (info->nodes * info->cores * info->threads);
You always have at least 1 numa node, at least 1 core, and at least 1 hyperthread, so you'd only get a zero in those if there was a bug in the Xen hypervisor Regards, Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
"Daniel P. Berrange" <berrange@redhat.com> wrote:
On Sat, Jan 19, 2008 at 09:21:07PM +0100, Jim Meyering wrote:
"Daniel P. Berrange" <berrange@redhat.com> wrote:
Xen 3.2.0 removed the sockets_per_node field from the nodeinfo data returned by XenD. I previously add compat for this, but got it in the wrong place so it would most likely end up in a divide-by-zero error. This patch re-arranges it to be correct.
diff -rup libvirt-0.4.0.orig/src/xend_internal.c libvirt-0.4.0.new/src/xend_internal.c --- libvirt-0.4.0.orig/src/xend_internal.c 2007-12-17 18:05:27.000000000 -0500 +++ libvirt-0.4.0.new/src/xend_internal.c 2008-01-18 21:13:30.000000000 -0500 @@ -1907,6 +1907,9 @@ sexpr_to_xend_node_info(struct sexpr *ro info->mhz = sexpr_int(root, "node/cpu_mhz"); info->nodes = sexpr_int(root, "node/nr_nodes"); info->sockets = sexpr_int(root, "node/sockets_per_node"); + info->cores = sexpr_int(root, "node/cores_per_socket"); + info->threads = sexpr_int(root, "node/threads_per_core"); + /* Xen 3.2.0 replaces sockets_per_node with 'nr_cpus'. * Old Xen calculated sockets_per_node using its internal * nr_cpus / (nodes*cores*threads), so fake it ourselves @@ -1921,8 +1924,6 @@ sexpr_to_xend_node_info(struct sexpr *ro if (info->sockets == 0) info->sockets = 1; } - info->cores = sexpr_int(root, "node/cores_per_socket"); - info->threads = sexpr_int(root, "node/threads_per_core"); return (0); }
ACK.
Now that you mention divide-by-zero, is there something that guarantees none of the terms in the denominator there is zero?
info->sockets = nr_cpus / (info->nodes * info->cores * info->threads);
You always have at least 1 numa node, at least 1 core, and at least 1 hyperthread, so you'd only get a zero in those if there was a bug in the Xen hypervisor
Yes, I understood that any of those being zero would not make sense. I mean, what if a caller of sexpr_to_xend_node_info were to provide a bogus root sexpr? Is it worth the added comparison to protect against that? int p = info->nodes * info->cores * info->threads; if (p == 0) return -1; info->sockets = nr_cpus / p;
Daniel P. Berrange wrote:
On Sat, Jan 19, 2008 at 09:21:07PM +0100, Jim Meyering wrote:
"Daniel P. Berrange" <berrange@redhat.com> wrote:
Xen 3.2.0 removed the sockets_per_node field from the nodeinfo data returned by XenD. I previously add compat for this, but got it in the wrong place so it would most likely end up in a divide-by-zero error. This patch re-arranges it to be correct.
diff -rup libvirt-0.4.0.orig/src/xend_internal.c libvirt-0.4.0.new/src/xend_internal.c --- libvirt-0.4.0.orig/src/xend_internal.c 2007-12-17 18:05:27.000000000 -0500 +++ libvirt-0.4.0.new/src/xend_internal.c 2008-01-18 21:13:30.000000000 -0500 @@ -1907,6 +1907,9 @@ sexpr_to_xend_node_info(struct sexpr *ro info->mhz = sexpr_int(root, "node/cpu_mhz"); info->nodes = sexpr_int(root, "node/nr_nodes"); info->sockets = sexpr_int(root, "node/sockets_per_node"); + info->cores = sexpr_int(root, "node/cores_per_socket"); + info->threads = sexpr_int(root, "node/threads_per_core"); + /* Xen 3.2.0 replaces sockets_per_node with 'nr_cpus'. * Old Xen calculated sockets_per_node using its internal * nr_cpus / (nodes*cores*threads), so fake it ourselves @@ -1921,8 +1924,6 @@ sexpr_to_xend_node_info(struct sexpr *ro if (info->sockets == 0) info->sockets = 1; } - info->cores = sexpr_int(root, "node/cores_per_socket"); - info->threads = sexpr_int(root, "node/threads_per_core"); return (0); } ACK.
Now that you mention divide-by-zero, is there something that guarantees none of the terms in the denominator there is zero?
info->sockets = nr_cpus / (info->nodes * info->cores * info->threads);
You always have at least 1 numa node, at least 1 core, and at least 1 hyperthread, so you'd only get a zero in those if there was a bug in the Xen hypervisor
.. And that would never happen :-) Rich. -- Emerging Technologies, Red Hat - http://et.redhat.com/~rjones/ Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE, United Kingdom. Registered in England and Wales under Company Registration No. 03798903
On Sun, Jan 20, 2008 at 12:22:49PM +0000, Richard W.M. Jones wrote:
Daniel P. Berrange wrote:
On Sat, Jan 19, 2008 at 09:21:07PM +0100, Jim Meyering wrote:
"Daniel P. Berrange" <berrange@redhat.com> wrote:
Xen 3.2.0 removed the sockets_per_node field from the nodeinfo data returned by XenD. I previously add compat for this, but got it in the wrong place so it would most likely end up in a divide-by-zero error. This patch re-arranges it to be correct.
diff -rup libvirt-0.4.0.orig/src/xend_internal.c libvirt-0.4.0.new/src/xend_internal.c --- libvirt-0.4.0.orig/src/xend_internal.c 2007-12-17 18:05:27.000000000 -0500 +++ libvirt-0.4.0.new/src/xend_internal.c 2008-01-18 21:13:30.000000000 -0500 @@ -1907,6 +1907,9 @@ sexpr_to_xend_node_info(struct sexpr *ro info->mhz = sexpr_int(root, "node/cpu_mhz"); info->nodes = sexpr_int(root, "node/nr_nodes"); info->sockets = sexpr_int(root, "node/sockets_per_node"); + info->cores = sexpr_int(root, "node/cores_per_socket"); + info->threads = sexpr_int(root, "node/threads_per_core"); + /* Xen 3.2.0 replaces sockets_per_node with 'nr_cpus'. * Old Xen calculated sockets_per_node using its internal * nr_cpus / (nodes*cores*threads), so fake it ourselves @@ -1921,8 +1924,6 @@ sexpr_to_xend_node_info(struct sexpr *ro if (info->sockets == 0) info->sockets = 1; } - info->cores = sexpr_int(root, "node/cores_per_socket"); - info->threads = sexpr_int(root, "node/threads_per_core"); return (0); } ACK.
Now that you mention divide-by-zero, is there something that guarantees none of the terms in the denominator there is zero?
info->sockets = nr_cpus / (info->nodes * info->cores * info->threads);
You always have at least 1 numa node, at least 1 core, and at least 1 hyperthread, so you'd only get a zero in those if there was a bug in the Xen hypervisor
.. And that would never happen :-)
Ok, ok :-) I applied the fix with Jim's suggested sanity check for 0 to protect against future Xen bugs :-) Regards, Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
participants (3)
-
Daniel P. Berrange -
Jim Meyering -
Richard W.M. Jones