tools: Skip backward time entries v2 (#1958)

* tools: Skip backward time entries in xfsslower

While using xfsslower on RHEL7 we occasionally get
following screwed up latencies:

  # xfsslower
  Tracing XFS operations slower than 1 ms
  TIME     COMM           PID    T BYTES   OFF_KB   LAT(ms) FILENAME
  13:25:03 git            3385   R 62      4704     18446744073708.55 tmp_pack_bDUbwZ
  13:25:03 git            3385   S 0       0           3.05 tmp_idx_Kjb2bW
  ...

The reason for this is that on RHEL7 it's possible to get backward
timetamp with bpf_ktime_get_ns. This needs to be fixed, but meanwhile
this fix makes sure the latencies with backward times are skipped.

For the rest of the kernels this is just sanity fix with
possibly just single compare instruction overhead.

It's temporary workaround for #728.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

* tools: Skip backward time entries in ext4dist

While using ext4dist on RHEL7 we occasionally get
following screwed up latencies:

  # ext4dist
  Tracing ext4 operation latency... Hit Ctrl-C to end.
  ^C

  operation = write
               usecs                         : count     distribution
                   0 -> 1                    : 1134529   |********            |
                   2 -> 3                    : 2777582   |********************|
                   4 -> 7                    : 688014    |****                |
                   8 -> 15                   : 36160     |                    |
                  16 -> 31                   : 698       |                    |
                  32 -> 63                   : 6         |                    |
                  64 -> 127                  : 15        |                    |
                 128 -> 255                  : 7         |                    |
                 256 -> 511                  : 1         |                    |
                 512 -> 1023                 : 0         |                    |
                1024 -> 2047                 : 0         |                    |
                2048 -> 4095                 : 2         |                    |
                4096 -> 8191                 : 1         |                    |
                8192 -> 16383                : 5         |                    |
               16384 -> 32767                : 0         |                    |
               32768 -> 65535                : 0         |                    |

    9007199254740992 -> 18014398509481983    : 0         |                    |
   18014398509481984 -> 36028797018963967    : 1         |                    |

The reason for this is that on RHEL7 it's possible to get backward
timestamp with bpf_ktime_get_ns. This needs to be fixed, but meanwhile
this fix makes sure the latencies with backward times are skipped.

For the rest of the kernels this is just sanity fix with
possibly just single compare instruction overhead.

It's temporary workaround for #728.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

tools/ext4dist.py

@@ -110,14 +110,20 @@ static int trace_return(struct pt_regs *ctx, const char *op)
     if (tsp == 0) {
         return 0;   // missed start or filtered
     }
-    u64 delta = (bpf_ktime_get_ns() - *tsp) / FACTOR;
+    u64 delta = bpf_ktime_get_ns() - *tsp;
+    start.delete(&pid);
+
+    // Skip entries with backwards time: temp workaround for #728
+    if ((s64) delta < 0)
+        return 0;
+
+    delta /= FACTOR;
 
     // store as histogram
     dist_key_t key = {.slot = bpf_log2l(delta)};
     __builtin_memcpy(&key.op, op, sizeof(key.op));
     dist.increment(key);
 
-    start.delete(&pid);
     return 0;
 }
 

tools/xfsslower.py

@@ -173,8 +173,15 @@ static int trace_return(struct pt_regs *ctx, int type)
 
     // calculate delta
     u64 ts = bpf_ktime_get_ns();
-    u64 delta_us = (ts - valp->ts) / 1000;
+    u64 delta_us = ts - valp->ts;
     entryinfo.delete(&id);
+
+    // Skip entries with backwards time: temp workaround for #728
+    if ((s64) delta_us < 0)
+        return 0;
+
+    delta_us /= 1000;
+
     if (FILTER_US)
         return 0;