[libvirt] [PATCH] rpc: allow truncated return for virDomainGetCPUStats
The RPC code assumed that the array returned by the driver would be fully populated; that is, ncpus on entry resulted in ncpus * return value on exit. However, while we don't support holes in the middle of ncpus, we do want to permit the case of ncpus on entry being longer than the array returned by the driver (that is, it should be safe for the caller to pass ncpus=128 on entry, and the driver will stop populating the array when it hits max_id). There are now three cases: server 0.9.10 and client 0.9.10 or newer: No impact - there were no hypervisor drivers that supported cpu stats server 0.9.11 or newer and client 0.9.10: if the client calls with ncpus beyond the max, then the rpc call will fail on the client side and disconnect the client, but the server is no worse for the wear server 0.9.11 or newer and client 0.9.11: the server can return a truncated array and the client will do just fine I reproduced the problem by using a host with 2 CPUs, and doing: virsh cpu-stats $dom --start 1 --count 2 * daemon/remote.c (remoteDispatchDomainGetCPUStats): Allow driver to omit tail of array. * src/remote/remote_driver.c (remoteDomainGetCPUStats): Accommodate driver that omits tail of array. --- daemon/remote.c | 10 ++++++++-- src/remote/remote_driver.c | 6 ++++-- 2 files changed, 12 insertions(+), 4 deletions(-) diff --git a/daemon/remote.c b/daemon/remote.c index 74a5f16..39302cc 100644 --- a/daemon/remote.c +++ b/daemon/remote.c @@ -3574,11 +3574,17 @@ remoteDispatchDomainGetCPUStats(virNetServerPtr server ATTRIBUTE_UNUSED, args->flags) < 0) goto cleanup; - percpu_len = ret->params.params_len / args->ncpus; - success: rv = 0; ret->nparams = percpu_len; + if (args->nparams && !(args->flags & VIR_TYPED_PARAM_STRING_OKAY)) { + int i; + + for (i = 0; i < percpu_len; i++) { + if (params[i].type == VIR_TYPED_PARAM_STRING) + ret->nparams--; + } + } cleanup: if (rv < 0) diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index 9e74cea..031167a 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -2382,7 +2382,7 @@ static int remoteDomainGetCPUStats(virDomainPtr domain, /* Check the length of the returned list carefully. */ if (ret.params.params_len > nparams * ncpus || (ret.params.params_len && - ret.nparams * ncpus != ret.params.params_len)) { + ((ret.params.params_len % ret.nparams) || ret.nparams > nparams))) { remoteError(VIR_ERR_RPC, "%s", _("remoteDomainGetCPUStats: " "returned number of stats exceeds limit")); @@ -2399,9 +2399,11 @@ static int remoteDomainGetCPUStats(virDomainPtr domain, } /* The remote side did not send back any zero entries, so we have - * to expand things back into a possibly sparse array. + * to expand things back into a possibly sparse array, where the + * tail of the array may be omitted. */ memset(params, 0, sizeof(*params) * nparams * ncpus); + ncpus = ret.params.params_len / ret.nparams; for (cpu = 0; cpu < ncpus; cpu++) { int tmp = nparams; remote_typed_param *stride = &ret.params.params_val[cpu * ret.nparams]; -- 1.7.7.6
On 03/07/2012 12:48 PM, Eric Blake wrote:
The RPC code assumed that the array returned by the driver would be fully populated; that is, ncpus on entry resulted in ncpus * return value on exit. However, while we don't support holes in the middle of ncpus, we do want to permit the case of ncpus on entry being longer than the array returned by the driver (that is, it should be safe for the caller to pass ncpus=128 on entry, and the driver will stop populating the array when it hits max_id).
There are now three cases: server 0.9.10 and client 0.9.10 or newer: No impact - there were no hypervisor drivers that supported cpu stats
server 0.9.11 or newer and client 0.9.10: if the client calls with ncpus beyond the max, then the rpc call will fail on the client side and disconnect the client, but the server is no worse for the wear
server 0.9.11 or newer and client 0.9.11: the server can return a truncated array and the client will do just fine
I reproduced the problem by using a host with 2 CPUs, and doing: virsh cpu-stats $dom --start 1 --count 2
* daemon/remote.c (remoteDispatchDomainGetCPUStats): Allow driver to omit tail of array. * src/remote/remote_driver.c (remoteDomainGetCPUStats): Accommodate driver that omits tail of array. --- daemon/remote.c | 10 ++++++++-- src/remote/remote_driver.c | 6 ++++-- 2 files changed, 12 insertions(+), 4 deletions(-)
diff --git a/daemon/remote.c b/daemon/remote.c index 74a5f16..39302cc 100644 --- a/daemon/remote.c +++ b/daemon/remote.c @@ -3574,11 +3574,17 @@ remoteDispatchDomainGetCPUStats(virNetServerPtr server ATTRIBUTE_UNUSED, args->flags)< 0) goto cleanup;
- percpu_len = ret->params.params_len / args->ncpus; - success: rv = 0; ret->nparams = percpu_len; + if (args->nparams&& !(args->flags& VIR_TYPED_PARAM_STRING_OKAY)) { + int i; + + for (i = 0; i< percpu_len; i++) { + if (params[i].type == VIR_TYPED_PARAM_STRING) + ret->nparams--; + } + }
cleanup: if (rv< 0) diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index 9e74cea..031167a 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -2382,7 +2382,7 @@ static int remoteDomainGetCPUStats(virDomainPtr domain, /* Check the length of the returned list carefully. */ if (ret.params.params_len> nparams * ncpus || (ret.params.params_len&& - ret.nparams * ncpus != ret.params.params_len)) { + ((ret.params.params_len % ret.nparams) || ret.nparams> nparams))) { remoteError(VIR_ERR_RPC, "%s", _("remoteDomainGetCPUStats: " "returned number of stats exceeds limit")); @@ -2399,9 +2399,11 @@ static int remoteDomainGetCPUStats(virDomainPtr domain, }
/* The remote side did not send back any zero entries, so we have - * to expand things back into a possibly sparse array. + * to expand things back into a possibly sparse array, where the + * tail of the array may be omitted. */ memset(params, 0, sizeof(*params) * nparams * ncpus); + ncpus = ret.params.params_len / ret.nparams; for (cpu = 0; cpu< ncpus; cpu++) { int tmp = nparams; remote_typed_param *stride =&ret.params.params_val[cpu * ret.nparams];
Make sense, and ACK. But do we want to add document to declare the returned array will be truncated among the API implementation. Not neccessary though. Osier
On 03/07/2012 02:46 PM, Osier Yang wrote:
On 03/07/2012 12:48 PM, Eric Blake wrote:
The RPC code assumed that the array returned by the driver would be fully populated; that is, ncpus on entry resulted in ncpus * return value on exit. However, while we don't support holes in the middle of ncpus, we do want to permit the case of ncpus on entry being longer than the array returned by the driver (that is, it should be safe for the caller to pass ncpus=128 on entry, and the driver will stop populating the array when it hits max_id).
There are now three cases: server 0.9.10 and client 0.9.10 or newer: No impact - there were no hypervisor drivers that supported cpu stats
server 0.9.11 or newer and client 0.9.10: if the client calls with ncpus beyond the max, then the rpc call will fail on the client side and disconnect the client, but the server is no worse for the wear
server 0.9.11 or newer and client 0.9.11: the server can return a truncated array and the client will do just fine
I reproduced the problem by using a host with 2 CPUs, and doing: virsh cpu-stats $dom --start 1 --count 2
* daemon/remote.c (remoteDispatchDomainGetCPUStats): Allow driver to omit tail of array. * src/remote/remote_driver.c (remoteDomainGetCPUStats): Accommodate driver that omits tail of array. --- daemon/remote.c | 10 ++++++++-- src/remote/remote_driver.c | 6 ++++-- 2 files changed, 12 insertions(+), 4 deletions(-)
diff --git a/daemon/remote.c b/daemon/remote.c index 74a5f16..39302cc 100644 --- a/daemon/remote.c +++ b/daemon/remote.c @@ -3574,11 +3574,17 @@ remoteDispatchDomainGetCPUStats(virNetServerPtr server ATTRIBUTE_UNUSED, args->flags)< 0) goto cleanup;
- percpu_len = ret->params.params_len / args->ncpus; - success: rv = 0; ret->nparams = percpu_len; + if (args->nparams&& !(args->flags& VIR_TYPED_PARAM_STRING_OKAY)) { + int i; + + for (i = 0; i< percpu_len; i++) { + if (params[i].type == VIR_TYPED_PARAM_STRING) + ret->nparams--; + } + }
cleanup: if (rv< 0) diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index 9e74cea..031167a 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -2382,7 +2382,7 @@ static int remoteDomainGetCPUStats(virDomainPtr domain, /* Check the length of the returned list carefully. */ if (ret.params.params_len> nparams * ncpus || (ret.params.params_len&& - ret.nparams * ncpus != ret.params.params_len)) { + ((ret.params.params_len % ret.nparams) || ret.nparams> nparams))) { remoteError(VIR_ERR_RPC, "%s", _("remoteDomainGetCPUStats: " "returned number of stats exceeds limit")); @@ -2399,9 +2399,11 @@ static int remoteDomainGetCPUStats(virDomainPtr domain, }
/* The remote side did not send back any zero entries, so we have - * to expand things back into a possibly sparse array. + * to expand things back into a possibly sparse array, where the + * tail of the array may be omitted. */ memset(params, 0, sizeof(*params) * nparams * ncpus); + ncpus = ret.params.params_len / ret.nparams; for (cpu = 0; cpu< ncpus; cpu++) { int tmp = nparams; remote_typed_param *stride =&ret.params.params_val[cpu * ret.nparams];
Make sense, and ACK.
But do we want to add document to declare the returned array will be truncated among the API implementation. Not neccessary though.
Perhaps something like: * whole). Otherwise, @start_cpu represents which cpu to start * with, and @ncpus represents how many consecutive processors to * query, with statistics attributable per processor (such as - * per-cpu usage). + * per-cpu usage). If @ncpus is larger than the number of host + * CPUs, the exceeded one(s) will be just ignored.
Osier
-- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On 03/07/2012 12:18 AM, Osier Yang wrote:
On 03/07/2012 02:46 PM, Osier Yang wrote:
On 03/07/2012 12:48 PM, Eric Blake wrote:
The RPC code assumed that the array returned by the driver would be fully populated; that is, ncpus on entry resulted in ncpus * return value on exit. However, while we don't support holes in the middle of ncpus, we do want to permit the case of ncpus on entry being longer than the array returned by the driver (that is, it should be safe for the caller to pass ncpus=128 on entry, and the driver will stop populating the array when it hits max_id).
/* The remote side did not send back any zero entries, so we have - * to expand things back into a possibly sparse array. + * to expand things back into a possibly sparse array, where the + * tail of the array may be omitted. */ memset(params, 0, sizeof(*params) * nparams * ncpus); + ncpus = ret.params.params_len / ret.nparams; for (cpu = 0; cpu< ncpus; cpu++) { int tmp = nparams; remote_typed_param *stride =&ret.params.params_val[cpu * ret.nparams];
Make sense, and ACK.
I realized that qemu_driver.c also needs this memset to guarantee that unused slots are returned empty on success.
But do we want to add document to declare the returned array will be truncated among the API implementation. Not neccessary though.
Perhaps something like:
* whole). Otherwise, @start_cpu represents which cpu to start * with, and @ncpus represents how many consecutive processors to * query, with statistics attributable per processor (such as - * per-cpu usage). + * per-cpu usage). If @ncpus is larger than the number of host + * CPUs, the exceeded one(s) will be just ignored.
Good idea. Here's what I squashed in before pushing: diff --git i/src/libvirt.c w/src/libvirt.c index d98741b..39ebb52 100644 --- i/src/libvirt.c +++ w/src/libvirt.c @@ -18534,7 +18534,9 @@ error: * whole). Otherwise, @start_cpu represents which cpu to start * with, and @ncpus represents how many consecutive processors to * query, with statistics attributable per processor (such as - * per-cpu usage). + * per-cpu usage). If @ncpus is larger than the number of cpus + * available to query, then the trailing part of the array will + * be unpopulated. * * The remote driver imposes a limit of 128 @ncpus and 16 @nparams; * the number of parameters per cpu should not exceed 16, but if you @@ -18579,8 +18581,10 @@ error: * number of populated @params, unless @ncpus was 1; and may be * less than @nparams). The populated parameters start at each * stride of @nparams, which means the results may be discontiguous; - * any unpopulated parameters will be zeroed on success. The caller - * is responsible for freeing any returned string parameters. + * any unpopulated parameters will be zeroed on success (this includes + * skipped elements if @nparams is too large, and tail elements if + * @ncpus is too large). The caller is responsible for freeing any + * returned string parameters. */ int virDomainGetCPUStats(virDomainPtr domain, virTypedParameterPtr params, diff --git i/src/qemu/qemu_driver.c w/src/qemu/qemu_driver.c index 538a419..ddb381a 100644 --- i/src/qemu/qemu_driver.c +++ w/src/qemu/qemu_driver.c @@ -12162,6 +12162,7 @@ qemuDomainGetPercpuStats(virDomainPtr domain, if (virCgroupGetCpuacctPercpuUsage(group, &buf)) goto cleanup; pos = buf; + memset(params, 0, nparams * ncpus); if (max_id - start_cpu > ncpus - 1) max_id = start_cpu + ncpus - 1; -- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
participants (2)
-
Eric Blake -
Osier Yang