[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] [PATCH v6 5/5] tools, docs: add total/local memory bandwith monitoring



Add Memory Bandwidth Monitoring(MBM) for VMs. Two types of monitoring
are supported: total and local memory bandwidth monitoring. To use it,
CMT should be enabled in hypervisor.

Signed-off-by: Chao Peng <chao.p.peng@xxxxxxxxxxxxxxx>
---
Changes in v6:
1. Remove DISABLE_IRQ flag as hypervisor disable IRQ for MSR_IA32_TSC
   implicitly.
Changes in v5:
1. Add MBM description in xen command line.
2. Use the tsc from hypervisor directly which is already ns.
3. Call resource_op with DISABLE_IRQ flag.
Changes in v4:
1. Get timestamp from hypervisor and use that for bandwidth calculation.
2. Minor document and coding style fix.
---
 docs/man/xl.pod.1                   |   9 ++++
 docs/misc/xen-command-line.markdown |   3 ++
 tools/libxc/include/xenctrl.h       |   5 +-
 tools/libxc/xc_msr_x86.h            |   1 +
 tools/libxc/xc_psr.c                |  35 +++++++++++--
 tools/libxl/libxl.h                 |   8 +++
 tools/libxl/libxl_psr.c             | 101 ++++++++++++++++++++++++++++++++++--
 tools/libxl/libxl_types.idl         |   2 +
 tools/libxl/xl_cmdimpl.c            |  21 +++++++-
 tools/libxl/xl_cmdtable.c           |   4 +-
 10 files changed, 178 insertions(+), 11 deletions(-)

diff --git a/docs/man/xl.pod.1 b/docs/man/xl.pod.1
index 6b89ba8..50759c0 100644
--- a/docs/man/xl.pod.1
+++ b/docs/man/xl.pod.1
@@ -1461,6 +1461,13 @@ is domain level. To monitor a specific domain, just 
attach the domain id with
 the monitoring service. When the domain doesn't need to be monitored any more,
 detach the domain id from the monitoring service.
 
+Intel Broadwell and later server platforms also offer total/local memory
+bandwidth monitoring. Xen supports per-domain monitoring for these two
+additional monitoring types. Both memory bandwidth monitoring and L3 cache
+occupancy monitoring share the same set of underlying monitoring service. Once
+a domain is attached to the monitoring service, monitoring data can be showed
+for any of these monitoring types.
+
 =over 4
 
 =item B<psr-cmt-attach> [I<domain-id>]
@@ -1476,6 +1483,8 @@ detach: Detach the platform shared resource monitoring 
service from a domain.
 Show monitoring data for a certain domain or all domains. Current supported
 monitor types are:
  - "cache-occupancy": showing the L3 cache occupancy.
+ - "total-mem-bandwidth": showing the total memory bandwidth.
+ - "local-mem-bandwidth": showing the local memory bandwidth.
 
 =back
 
diff --git a/docs/misc/xen-command-line.markdown 
b/docs/misc/xen-command-line.markdown
index a061aa4..0491dbb 100644
--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -1097,6 +1097,9 @@ The following resources are available:
   L3 cache occupancy.
   * `cmt` instructs Xen to enable/disable Cache Monitoring Technology.
   * `rmid_max` indicates the max value for rmid.
+* Memory Bandwidth Monitoring (Broadwell and later). Information regarding the
+  total/local memory bandwidth. Follow the same options with Cache Monitoring
+  Technology.
 
 ### reboot
 > `= t[riple] | k[bd] | a[cpi] | p[ci] | n[o] [, [w]arm | [c]old]`
diff --git a/tools/libxc/include/xenctrl.h b/tools/libxc/include/xenctrl.h
index 8a0eab6..8f964ad 100644
--- a/tools/libxc/include/xenctrl.h
+++ b/tools/libxc/include/xenctrl.h
@@ -2688,6 +2688,8 @@ int xc_resource_op(xc_interface *xch, uint32_t nr_ops, 
xc_resource_op_t *ops);
 #if defined(__i386__) || defined(__x86_64__)
 enum xc_psr_cmt_type {
     XC_PSR_CMT_L3_OCCUPANCY,
+    XC_PSR_CMT_TOTAL_MEM_BANDWIDTH,
+    XC_PSR_CMT_LOCAL_MEM_BANDWIDTH,
 };
 typedef enum xc_psr_cmt_type xc_psr_cmt_type;
 int xc_psr_cmt_attach(xc_interface *xch, uint32_t domid);
@@ -2701,7 +2703,8 @@ int xc_psr_cmt_get_l3_event_mask(xc_interface *xch, 
uint32_t *event_mask);
 int xc_psr_cmt_get_l3_cache_size(xc_interface *xch, uint32_t cpu,
                                  uint32_t *l3_cache_size);
 int xc_psr_cmt_get_data(xc_interface *xch, uint32_t rmid, uint32_t cpu,
-                        uint32_t psr_cmt_type, uint64_t *monitor_data);
+                        uint32_t psr_cmt_type, uint64_t *monitor_data,
+                        uint64_t *tsc);
 int xc_psr_cmt_enabled(xc_interface *xch);
 #endif
 
diff --git a/tools/libxc/xc_msr_x86.h b/tools/libxc/xc_msr_x86.h
index 7c3e1a3..7f100e7 100644
--- a/tools/libxc/xc_msr_x86.h
+++ b/tools/libxc/xc_msr_x86.h
@@ -20,6 +20,7 @@
 #ifndef XC_MSR_X86_H
 #define XC_MSR_X86_H
 
+#define MSR_IA32_TSC            0x00000010
 #define MSR_IA32_CMT_EVTSEL     0x00000c8d
 #define MSR_IA32_CMT_CTR        0x00000c8e
 
diff --git a/tools/libxc/xc_psr.c b/tools/libxc/xc_psr.c
index a9881a4..bf9ec72 100644
--- a/tools/libxc/xc_psr.c
+++ b/tools/libxc/xc_psr.c
@@ -23,6 +23,8 @@
 #define IA32_CMT_CTR_ERROR_MASK         (0x3ull << 62)
 
 #define EVTID_L3_OCCUPANCY             0x1
+#define EVTID_TOTAL_MEM_BANDWIDTH      0x2
+#define EVTID_LOCAL_MEM_BANDWIDTH      0x3
 
 int xc_psr_cmt_attach(xc_interface *xch, uint32_t domid)
 {
@@ -156,11 +158,12 @@ int xc_psr_cmt_get_l3_cache_size(xc_interface *xch, 
uint32_t cpu,
 }
 
 int xc_psr_cmt_get_data(xc_interface *xch, uint32_t rmid, uint32_t cpu,
-                        xc_psr_cmt_type type, uint64_t *monitor_data)
+                        xc_psr_cmt_type type, uint64_t *monitor_data,
+                        uint64_t *tsc)
 {
     xc_resource_op_t op;
-    xc_resource_entry_t entries[2];
-    uint32_t evtid;
+    xc_resource_entry_t entries[3];
+    uint32_t evtid, nr_entries;
     int rc;
 
     switch ( type )
@@ -168,6 +171,12 @@ int xc_psr_cmt_get_data(xc_interface *xch, uint32_t rmid, 
uint32_t cpu,
     case XC_PSR_CMT_L3_OCCUPANCY:
         evtid = EVTID_L3_OCCUPANCY;
         break;
+    case XC_PSR_CMT_TOTAL_MEM_BANDWIDTH:
+        evtid = EVTID_TOTAL_MEM_BANDWIDTH;
+        break;
+    case XC_PSR_CMT_LOCAL_MEM_BANDWIDTH:
+        evtid = EVTID_LOCAL_MEM_BANDWIDTH;
+        break;
     default:
         return -1;
     }
@@ -182,19 +191,35 @@ int xc_psr_cmt_get_data(xc_interface *xch, uint32_t rmid, 
uint32_t cpu,
     entries[1].val = 0;
     entries[1].rsvd = 0;
 
+    if ( type == XC_PSR_CMT_L3_OCCUPANCY )
+        nr_entries = 2;
+    else
+    {
+        entries[2].u.cmd = XEN_RESOURCE_OP_MSR_READ;
+        entries[2].idx = MSR_IA32_TSC;
+        entries[2].val = 0;
+        entries[2].rsvd = 0;
+
+        nr_entries = 3;
+    }
+
     op.cpu = cpu;
-    op.nr_entries = 2;
+    op.nr_entries = nr_entries;
     op.entries = entries;
 
     rc = xc_resource_op(xch, 1, &op);
     if ( rc < 0 )
         return rc;
 
-    if ( op.result !=2 || entries[1].val & IA32_CMT_CTR_ERROR_MASK )
+    if ( op.result != nr_entries || entries[1].val & IA32_CMT_CTR_ERROR_MASK )
         return -1;
 
     *monitor_data = entries[1].val;
 
+    if ( type == XC_PSR_CMT_TOTAL_MEM_BANDWIDTH ||
+         type == XC_PSR_CMT_LOCAL_MEM_BANDWIDTH )
+        *tsc = entries[2].val;
+
     return 0;
 }
 
diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
index 596d2a0..347ef52 100644
--- a/tools/libxl/libxl.h
+++ b/tools/libxl/libxl.h
@@ -1462,6 +1462,14 @@ int libxl_psr_cmt_get_cache_occupancy(libxl_ctx *ctx,
                                       uint32_t domid,
                                       uint32_t socketid,
                                       uint32_t *l3_cache_occupancy);
+int libxl_psr_cmt_get_total_mem_bandwidth(libxl_ctx *ctx,
+                                          uint32_t domid,
+                                          uint32_t socketid,
+                                          uint32_t *bandwidth);
+int libxl_psr_cmt_get_local_mem_bandwidth(libxl_ctx *ctx,
+                                          uint32_t domid,
+                                          uint32_t socketid,
+                                          uint32_t *bandwidth);
 #endif
 
 /* misc */
diff --git a/tools/libxl/libxl_psr.c b/tools/libxl/libxl_psr.c
index c88c421..4c8bcc8 100644
--- a/tools/libxl/libxl_psr.c
+++ b/tools/libxl/libxl_psr.c
@@ -18,6 +18,7 @@
 
 
 #define IA32_QM_CTR_ERROR_MASK         (0x3ul << 62)
+#define MBM_SAMPLE_RETRY_MAX 4
 
 static void libxl__psr_cmt_log_err_msg(libxl__gc *gc, int err)
 {
@@ -180,7 +181,8 @@ static int libxl__psr_cmt_get_l3_monitoring_data(libxl__gc 
*gc,
                                                  uint32_t domid,
                                                  xc_psr_cmt_type type,
                                                  uint32_t socketid,
-                                                 uint64_t *data)
+                                                 uint64_t *data,
+                                                 uint64_t *tsc)
 {
     unsigned int rmid;
     int cpu, rc;
@@ -200,7 +202,7 @@ static int libxl__psr_cmt_get_l3_monitoring_data(libxl__gc 
*gc,
         goto out;
     }
 
-    rc = xc_psr_cmt_get_data(CTX->xch, rmid, cpu, type, data);
+    rc = xc_psr_cmt_get_data(CTX->xch, rmid, cpu, type, data, tsc);
     if (rc < 0) {
         LOGE(ERROR, "failed to get monitoring data");
         rc = ERROR_FAIL;
@@ -222,7 +224,7 @@ int libxl_psr_cmt_get_cache_occupancy(libxl_ctx *ctx,
 
     rc = libxl__psr_cmt_get_l3_monitoring_data(gc, domid,
                                                XC_PSR_CMT_L3_OCCUPANCY,
-                                               socketid, &data);
+                                               socketid, &data, NULL);
     if (rc < 0)
             goto out;
 
@@ -240,6 +242,99 @@ out:
     return rc;
 }
 
+static int libxl__psr_cmt_get_mem_bandwidth(libxl__gc *gc,
+                                            uint32_t domid,
+                                            xc_psr_cmt_type type,
+                                            uint32_t socketid,
+                                            uint32_t *bandwidth)
+{
+    uint64_t sample1, sample2;
+    uint64_t tsc1, tsc2;
+    uint32_t upscaling_factor;
+    int retry_attempts = 0;
+    int rc;
+
+    while (1) {
+        rc = libxl__psr_cmt_get_l3_monitoring_data(gc, domid, type, socketid,
+                                                   &sample1, &tsc1);
+        if (rc < 0) {
+            rc = ERROR_FAIL;
+            goto out;
+        }
+
+        usleep(10000);
+
+        rc = libxl__psr_cmt_get_l3_monitoring_data(gc, domid, type, socketid,
+                                                   &sample2, &tsc2);
+        if (rc < 0) {
+           rc = ERROR_FAIL;
+           goto out;
+        }
+        if (tsc2 <= tsc1) {
+            rc = ERROR_FAIL;
+            goto out;
+        }
+        /*
+         * Hardware guarantees at most 1 overflow can happen if the duration
+         * between two samples is less than 1 second. Note that tsc returned
+         * from hypervisor is already-scaled time(ns).
+         */
+        if (tsc2 - tsc1 < 1000000000 && sample2 >= sample1)
+            break;
+
+        if (retry_attempts < MBM_SAMPLE_RETRY_MAX) {
+            retry_attempts++;
+        } else {
+            LOGE(ERROR, "event counter overflowed");
+            rc = ERROR_FAIL;
+            goto out;
+        }
+
+    }
+
+    rc = xc_psr_cmt_get_l3_upscaling_factor(CTX->xch, &upscaling_factor);
+    if (rc < 0) {
+        LOGE(ERROR, "failed to get L3 upscaling factor");
+        rc = ERROR_FAIL;
+        goto out;
+    }
+
+    *bandwidth = (sample2 - sample1) * 1000000000 / (tsc2 - tsc1)
+                 *  upscaling_factor / 1024;
+out:
+    return rc;
+}
+
+int libxl_psr_cmt_get_total_mem_bandwidth(libxl_ctx *ctx,
+                                          uint32_t domid,
+                                          uint32_t socketid,
+                                          uint32_t *bandwidth)
+{
+    GC_INIT(ctx);
+    int rc;
+
+    rc = libxl__psr_cmt_get_mem_bandwidth(gc, domid,
+                                          XC_PSR_CMT_TOTAL_MEM_BANDWIDTH,
+                                          socketid, bandwidth);
+    GC_FREE;
+    return rc;
+}
+
+int libxl_psr_cmt_get_local_mem_bandwidth(libxl_ctx *ctx,
+                                          uint32_t domid,
+                                          uint32_t socketid,
+                                          uint32_t *bandwidth)
+{
+    GC_INIT(ctx);
+    int rc;
+
+    rc = libxl__psr_cmt_get_mem_bandwidth(gc, domid,
+                                          XC_PSR_CMT_LOCAL_MEM_BANDWIDTH,
+                                          socketid, bandwidth);
+    GC_FREE;
+    return rc;
+}
+
 /*
  * Local variables:
  * mode: C
diff --git a/tools/libxl/libxl_types.idl b/tools/libxl/libxl_types.idl
index 1214d2e..46d160e 100644
--- a/tools/libxl/libxl_types.idl
+++ b/tools/libxl/libxl_types.idl
@@ -694,4 +694,6 @@ libxl_event = Struct("event",[
 
 libxl_psr_cmt_type = Enumeration("psr_cmt_type", [
     (1, "CACHE_OCCUPANCY"),
+    (2, "TOTAL_MEM_BANDWIDTH"),
+    (3, "LOCAL_MEM_BANDWIDTH"),
     ])
diff --git a/tools/libxl/xl_cmdimpl.c b/tools/libxl/xl_cmdimpl.c
index 1827c63..6d5c7af 100644
--- a/tools/libxl/xl_cmdimpl.c
+++ b/tools/libxl/xl_cmdimpl.c
@@ -7829,6 +7829,16 @@ static void psr_cmt_print_domain_l3_info(libxl_dominfo 
*dominfo,
                                                    socketid, &data))
                 printf("%13u KB", data);
             break;
+        case LIBXL_PSR_CMT_TYPE_TOTAL_MEM_BANDWIDTH:
+            if (!libxl_psr_cmt_get_total_mem_bandwidth(ctx, dominfo->domid,
+                                                       socketid, &data))
+                printf("%11u KB/s", data);
+            break;
+        case LIBXL_PSR_CMT_TYPE_LOCAL_MEM_BANDWIDTH:
+            if (!libxl_psr_cmt_get_local_mem_bandwidth(ctx, dominfo->domid,
+                                                       socketid, &data))
+                printf("%11u KB/s", data);
+            break;
         default:
             return;
         }
@@ -7840,7 +7850,7 @@ static void psr_cmt_print_domain_l3_info(libxl_dominfo 
*dominfo,
 static int psr_cmt_show_l3_info(libxl_psr_cmt_type type, uint32_t domid)
 {
     uint32_t i, socketid, nr_sockets, total_rmid;
-    uint32_t l3_cache_size;
+    uint32_t l3_cache_size, l3_event_mask;
     libxl_physinfo info;
     int rc, nr_domains;
 
@@ -7849,6 +7859,13 @@ static int psr_cmt_show_l3_info(libxl_psr_cmt_type type, 
uint32_t domid)
         return -1;
     }
 
+    rc = libxl_psr_cmt_get_l3_event_mask(ctx, &l3_event_mask);
+    if (rc < 0 || !(l3_event_mask & (1 << (type - 1)))) {
+        fprintf(stderr, "Monitor type '%s' is not supported in the system\n",
+                libxl_psr_cmt_type_to_string(type));
+        return -1;
+    }
+
     libxl_physinfo_init(&info);
     rc = libxl_get_physinfo(ctx, &info);
     if (rc < 0) {
@@ -7966,6 +7983,8 @@ int main_psr_cmt_show(int argc, char **argv)
 
     switch (type) {
     case LIBXL_PSR_CMT_TYPE_CACHE_OCCUPANCY:
+    case LIBXL_PSR_CMT_TYPE_TOTAL_MEM_BANDWIDTH:
+    case LIBXL_PSR_CMT_TYPE_LOCAL_MEM_BANDWIDTH:
         ret = psr_cmt_show_l3_info(type, domid);
         break;
     default:
diff --git a/tools/libxl/xl_cmdtable.c b/tools/libxl/xl_cmdtable.c
index 4b30d3d..2d8f272 100644
--- a/tools/libxl/xl_cmdtable.c
+++ b/tools/libxl/xl_cmdtable.c
@@ -538,7 +538,9 @@ struct cmd_spec cmd_table[] = {
       "Show Cache Monitoring Technology information",
       "<PSR-CMT-Type> <Domain>",
       "Available monitor types:\n"
-      "\"cache_occupancy\":         Show L3 cache occupancy\n",
+      "\"cache_occupancy\":         Show L3 cache occupancy\n"
+      "\"total_mem_bandwidth\":     Show total memory bandwidth\n"
+      "\"local_mem_bandwidth\":     Show local memory bandwidth\n",
     },
 #endif
 };
-- 
1.9.1


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.