test: Allow aggpercpu test to omit some CPUs
PATCH 1/8 test: Allow aggpercpu test to omit some CPUs
From: Eugene Loh <[email protected]>
The aggpercpu test uses profile-* to fire aggregations on every CPU. But whether for good (offline CPUs) or bad (unreliable probe) reasons, a CPU might not aggregate any values.
Allow this test to skip over such CPUs. That is, while a BPF agg map might have the correct default value for the aggregation function, it may not see any anticipated data. The quality of profile-* probes should be checked by a different test.
Also, when computing avg() and stddev() in the awk check program, truncate results to ints more often to mimic the algorithms in DTrace for more robust checks.
--- test/unittest/aggs/tst.aggpercpu.sh | 62 +++++++++++++++++++++-------- 1 file changed, 45 insertions(+), 17 deletions(-)
test/unittest/aggs/tst.aggpercpu.sh +45 -17
@@ -46,7 +46,8 @@ fi awk ' # The expected value for the aggregation is aggval. # The expected value on a CPU is (m * cpu + b). - function check(label, aggval, m, b) { + # The default value on a CPU that did not fire is defval. + function check(label, aggval, m, b, defval) { # Check the aggregation over all CPUs. getline; print "check:", $0;
@@ -54,12 +55,38 @@ awk ' # Check the per-CPU values. for (i = 1; i <= ncpu; i++) { - getline; - print "check:", $0; - if (match($0, "^ \\[CPU ") != 1 || - strtonum($2) != cpu[i] || - strtonum($3) != m * cpu[i] + b) - printf("ERROR: %s, agg per cpu %d, line: %s\n", label, cpu[i], $0); + do { + getline; + print "check:", $0; + + # pass is +1 for pass, -1 for error, 0 for skip CPU + pass = 0; + + if (match($0, "^ \\[CPU ") != 1) { + print "ERROR: no CPU to read" + pass = -1; + } else if (strtonum($2) > cpu[i]) { + print "ERROR: skipped over expected CPU" + pass = -1; + } else if (strtonum($2) == cpu[i]) { + if (strtonum($3) != m * cpu[i] + b) { + print "ERROR: wrong value" + pass = -1; + } else { + # right value + pass = +1; + } + } else if ($3 != defval) { + print "ERROR: wrong default value" + pass = -1; + } else { + # skip over CPU that apparently did not fire + pass = 0; + } + + if (pass == -1) + printf("ERROR: %s, agg per cpu %d, line: %s\n", label, cpu[i], $0); + } while (pass == 0); } }
@@ -99,13 +126,14 @@ awk ' { # First we finish computing our estimates for avg and stddev. # (The other results require no further action.) + # (We keep truncating to ints to mimic DTrace algorithms.) - xavg /= xcnt; + xavg /= xcnt; xavg = int(xavg); - xstm /= xcnt; - xstd /= xcnt; + xstm /= xcnt; xstm = int(xstm); + xstd /= xcnt; xstd = int(xstd); xstd -= xstm * xstm; - xstd = int(sqrt(xstd)); + xstd = sqrt(xstd); xstd = int(xstd); # Sort the cpus.
@@ -113,12 +141,12 @@ awk ' # Now read the results and compare. - check("cnt", xcnt, 0, 1); - check("avg", xavg, 10, 3); - check("std", xstd, 0, 0); - check("min", xmin, 30, -10); - check("max", xmax, 40, -15); - check("sum", xsum, 50, 0); + check("cnt", xcnt, 0, 1, "0"); + check("avg", xavg, 10, 3, "0"); + check("std", xstd, 0, 0, "0"); + check("min", xmin, 30, -10, "9223372036854775807"); + check("max", xmax, 40, -15, "-9223372036854775808"); + check("sum", xsum, 50, 0, "0"); printf("done\n"); }
Signature · 1 line
2.18.4
PATCH 2/8 Reduce stack depth if kernel returns NULL frames
From: Eugene Loh <[email protected]>
The return value from the BPF helper function bpf_get_stack() basically returns the size of the stack returned. We use this value to report stack depth.
Some of the top frames can be NULL, however, leading to some inconsistencies between reported stacks and stack depths.
Add some code to reduce the stack depth if one or two top frames are NULL.
There is an existing test to check for this problem. It will appear in a later patch since it has multiple problems.
--- bpf/get_bvar.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)
bpf/get_bvar.c +15 -2
@@ -67,7 +67,9 @@ noinline uint64_t dt_get_bvar(const dt_dctx_t *dctx, uint32_t id, uint32_t idx)uint32_t bufsiz = (uint32_t ) (uint64_t ) (&STKSIZ );uint64_t flags;char *buf = dctx->mem + (uint64_t )(&STACK_OFF ); -uint64_t stacksize; +int64_t stacksize; +int64_t topslot; +uint64_t *pcs = (uint64_t *)buf;if (id ==DIF_VAR_USTACKDEPTH ) flags =BPF_F_USER_STACK ;
@@ -87,8 +89,19 @@ noinline uint64_t dt_get_bvar(const dt_dctx_t *dctx, uint32_t id, uint32_t idx)* what we can retrieve. But it's also possible that the * buffer was exactly large enough. So, leave it to the user * to interpret the result. +* +* The helper function also sometimes returns some empty frames +* at the top. Bump the depth down some so that the stack depth +* we report is consistent with the number of frames returned. +* Arguably, this should be fixed in the kernel, but we can +* work around the problem for now. */ -return stacksize /sizeof (uint64_t ); + topslot = stacksize /sizeof (uint64_t ) -1 ; +if (topslot >=0 && topslot < (bufsiz /sizeof (uint64_t )) && pcs[topslot] ==0 ) + topslot--; +if (topslot >=0 && topslot < (bufsiz /sizeof (uint64_t )) && pcs[topslot] ==0 ) + topslot--; +return topslot +1 ; }case DIF_VAR_CALLER :case DIF_VAR_UCALLER : {
Signature · 1 line
2.18.4
Re: [DTrace-devel] [PATCH 2/8] Reduce stack depth if kernel returns NULL frames
If bpf_get_stack() can give us stacks with NULL pointers at the top, wouldn't we need code to remove those (if they need to be removed) from the actual stack() data also? If they are left there, then I would argue that we should also include them in the stackdepth count.
Quoted 64 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
The return value from the BPF helper function
bpf_get_stack()basically returns the size of the stack returned. We use this value to report stack depth.Some of the top frames can be NULL, however, leading to some inconsistencies between reported stacks and stack depths.
Add some code to reduce the stack depth if one or two top frames are NULL.
There is an existing test to check for this problem. It will appear in a later patch since it has multiple problems.
Signed-off-by Eugene Loh <[email protected]>--- bpf/get_bvar.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)bpf/get_bvar.c +15
-2@@ -67,7 +67,9 @@ noinline uint64_t dt_get_bvar(const dt_dctx_t *dctx, uint32_t id, uint32_t idx)uint32_t bufsiz = (uint32_t ) (uint64_t ) (&STKSIZ );uint64_t flags;char *buf = dctx->mem + (uint64_t )(&STACK_OFF ); -uint64_t stacksize; +int64_t stacksize; +int64_t topslot; +uint64_t *pcs = (uint64_t *)buf;if (id ==DIF_VAR_USTACKDEPTH ) flags =BPF_F_USER_STACK ;@@ -87,8 +89,19 @@ noinline uint64_t dt_get_bvar(const dt_dctx_t *dctx, uint32_t id, uint32_t idx)* what we can retrieve. But it's also possible that the * buffer was exactly large enough. So, leave it to the user * to interpret the result. +* +* The helper function also sometimes returns some empty frames +* at the top. Bump the depth down some so that the stack depth +* we report is consistent with the number of frames returned. +* Arguably, this should be fixed in the kernel, but we can +* work around the problem for now. */ -return stacksize /sizeof (uint64_t ); + topslot = stacksize /sizeof (uint64_t ) -1 ; +if (topslot >=0 && topslot < (bufsiz /sizeof (uint64_t )) && pcs[topslot] ==0 ) + topslot--; +if (topslot >=0 && topslot < (bufsiz /sizeof (uint64_t )) && pcs[topslot] ==0 ) + topslot--; +return topslot +1 ; }case DIF_VAR_CALLER :case DIF_VAR_UCALLER : {Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
Re: [DTrace-devel] [PATCH 2/8] Reduce stack depth if kernel returns NULL frames
It's been a while, but if I remember correctly it's actually just the other way around. That is, to make stack and depth consistent, we're forced into this "depth reduction" patch. Put another way, the stack simply does not include NULL pointers -- there is no way to remove them.
E.g., let's say we have a buffer of 8 pointers and we ask for the stack and get back:
0xdead 0xbeef 0xfeed 0xface NULL NULL NULL NULL
Looks like 4 pointers. But we don't count them. We just use the return value from the helper function. If it tells us "4", then everything is consistent. But for some reason (I haven't looked at the code to figure out why), it can be 5 or even 6. So this patch bumps that value down -- to attain the consistency I think you're asking about.
There is a test that exposes the problem (inconsistency between stack and depth) and this patch's fix. Again, the stack already does not have NULL pointers, and so it's simply a matter of patching up the depth. Unfortunately, the test is riddled with all sorts of problems and I haven't yet gotten around to cleaning it up to the point where I can put it back.
Quoted 70 lines · Kris Van Hees
If
bpf_get_stack()can give us stacks with NULL pointers at the top, wouldn't we need code to remove those (if they need to be removed) from the actualstack()data also? If they are left there, then I would argue that we should also include them in the stackdepth count.Quoted 64 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
The return value from the BPF helper function
bpf_get_stack()basically returns the size of the stack returned. We use this value to report stack depth.Some of the top frames can be NULL, however, leading to some inconsistencies between reported stacks and stack depths.
Add some code to reduce the stack depth if one or two top frames are NULL.
There is an existing test to check for this problem. It will appear in a later patch since it has multiple problems.
Signed-off-by Eugene Loh <[email protected]>--- bpf/get_bvar.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)bpf/get_bvar.c +15
-2@@ -67,7 +67,9 @@ noinline uint64_t dt_get_bvar(const dt_dctx_t *dctx, uint32_t id, uint32_t idx)uint32_t bufsiz = (uint32_t ) (uint64_t ) (&STKSIZ );uint64_t flags;char *buf = dctx->mem + (uint64_t )(&STACK_OFF ); -uint64_t stacksize; +int64_t stacksize; +int64_t topslot; +uint64_t *pcs = (uint64_t *)buf;if (id ==DIF_VAR_USTACKDEPTH ) flags =BPF_F_USER_STACK ;@@ -87,8 +89,19 @@ noinline uint64_t dt_get_bvar(const dt_dctx_t *dctx, uint32_t id, uint32_t idx)* what we can retrieve. But it's also possible that the * buffer was exactly large enough. So, leave it to the user * to interpret the result. +* +* The helper function also sometimes returns some empty frames +* at the top. Bump the depth down some so that the stack depth +* we report is consistent with the number of frames returned. +* Arguably, this should be fixed in the kernel, but we can +* work around the problem for now. */ -return stacksize /sizeof (uint64_t ); + topslot = stacksize /sizeof (uint64_t ) -1 ; +if (topslot >=0 && topslot < (bufsiz /sizeof (uint64_t )) && pcs[topslot] ==0 ) + topslot--; +if (topslot >=0 && topslot < (bufsiz /sizeof (uint64_t )) && pcs[topslot] ==0 ) + topslot--; +return topslot +1 ; }case DIF_VAR_CALLER :case DIF_VAR_UCALLER : {Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
Re: [DTrace-devel] [PATCH 2/8] Reduce stack depth if kernel returns NULL frames
Quoted 8 lines · Eugene Loh
It's been a while, but if I remember correctly it's actually just the other way around. That is, to make stack and depth consistent, we're forced into this "depth reduction" patch. Put another way, the stack simply does not include NULL pointers -- there is no way to remove them.
E.g., let's say we have a buffer of 8 pointers and we ask for the stack and get back:
Quoted 7 lines
0xdead0xbeef0xfeed0xfaceNULL NULL NULL NULLLooks like 4 pointers. But we don't count them. We just use the return value from the helper function. If it tells us "4", then everything is consistent. But for some reason (I haven't looked at the code to figure out why), it can be 5 or even 6. So this patch bumps that value down -- to attain the consistency I think you're asking about.
So you are saying that the number of values that is actually filled in is not consistent with the return value of the bpf_get_stack() helper? That would sound like a kernel bug.
Re: [DTrace-devel] [PATCH 2/8] Reduce stack depth if kernel returns NULL frames
Quoted 11 lines · Kris Van Hees
Quoted 10 lines · Eugene Loh
It's been a while, but if I remember correctly it's actually just the other way around. That is, to make stack and depth consistent, we're forced into this "depth reduction" patch. Put another way, the stack simply does not include NULL pointers -- there is no way to remove them.
E.g., let's say we have a buffer of 8 pointers and we ask for the stack and get back:
0xdead0xbeef0xfeed0xfaceNULL NULL NULL NULL
Quoted 8 lines
Quoted 5 lines
Looks like 4 pointers. But we don't count them. We just use the return value from the helper function. If it tells us "4", then everything is consistent. But for some reason (I haven't looked at the code to figure out why), it can be 5 or even 6. So this patch bumps that value down -- to attain the consistency I think you're asking about.
So you are saying that the number of values that is actually filled in is not consistent with the return value of the
bpf_get_stack()helper? That would sound like a kernel bug.
I think there might be a kernel bug. For a variety of practical reasons, I went with a DTrace workaround (and haven't looked at the kernel code).
Re: [DTrace-devel] [PATCH 2/8] Reduce stack depth if kernel returns NULL frames
Quoted 18 lines · Eugene Loh via DTrace-devel
Quoted 16 lines · Kris Van Hees
Quoted 15 lines · Eugene Loh
It's been a while, but if I remember correctly it's actually just the other way around. That is, to make stack and depth consistent, we're forced into this "depth reduction" patch. Put another way, the stack simply does not include NULL pointers -- there is no way to remove them.
E.g., let's say we have a buffer of 8 pointers and we ask for the stack and get back:
0xdead0xbeef0xfeed0xfaceNULL NULL NULL NULLLooks like 4 pointers. But we don't count them. We just use the return value from the helper function. If it tells us "4", then everything is consistent. But for some reason (I haven't looked at the code to figure out why), it can be 5 or even 6. So this patch bumps that value down -- to attain the consistency I think you're asking about.
Quoted 5 lines
Quoted 3 lines
So you are saying that the number of values that is actually filled in is not consistent with the return value of the
bpf_get_stack()helper? That would sound like a kernel bug.I think there might be a kernel bug. For a variety of practical reasons, I went with a DTrace workaround (and haven't looked at the kernel code).
I'll look around to see if I can figure out what is going on and whether it is a kernel bug (which of course may already have been fixed in some kernel version). In general, we should mention in patches like this that it seems to be a kernel bug (or first determine if it is) because otherwise we're going to implement workarounds without ever making sure the kernel bug gets fixed. Or if it is not a kernel bug, we need to look elsewhere why this goes wrong.
Re: [DTrace-devel] [PATCH 2/8] Reduce stack depth if kernel returns NULL frames
I cannot remember which test this patch was trying to fix and recent test results do not suggest any obvious candidates. Under the circumstances, I'm fine rescinding this patch. Arguably it was just a workaround for a kernel bug anyhow, and perhaps the underlying kernel bug has been fixed.
Quoted 29 lines · Kris Van Hees
Quoted 18 lines · Eugene Loh via DTrace-devel
Quoted 16 lines · Kris Van Hees
Quoted 15 lines · Eugene Loh
It's been a while, but if I remember correctly it's actually just the other way around. That is, to make stack and depth consistent, we're forced into this "depth reduction" patch. Put another way, the stack simply does not include NULL pointers -- there is no way to remove them.
E.g., let's say we have a buffer of 8 pointers and we ask for the stack and get back:
0xdead0xbeef0xfeed0xfaceNULL NULL NULL NULLLooks like 4 pointers. But we don't count them. We just use the return value from the helper function. If it tells us "4", then everything is consistent. But for some reason (I haven't looked at the code to figure out why), it can be 5 or even 6. So this patch bumps that value down -- to attain the consistency I think you're asking about.
Quoted 4 lines
Quoted 3 lines
So you are saying that the number of values that is actually filled in is not consistent with the return value of the
bpf_get_stack()helper? That would sound like a kernel bug.I think there might be a kernel bug. For a variety of practical reasons, I went with a DTrace workaround (and haven't looked at the kernel code).
I'll look around to see if I can figure out what is going on and whether it is a kernel bug (which of course may already have been fixed in some kernel version). In general, we should mention in patches like this that it seems to be a kernel bug (or first determine if it is) because otherwise we're going to implement workarounds without ever making sure the kernel bug gets fixed. Or if it is not a kernel bug, we need to look elsewhere why this goes wrong.
PATCH 3/8 test: Fix nonexistent @@sort option
In reply to [email protected]From: Eugene Loh <[email protected]>
--- test/unittest/options/tst.aggrate-fast.d | 4 ++-- test/unittest/options/tst.aggrate-fast.r | 2 +- test/unittest/options/tst.aggrate-slow.d | 4 ++-- test/unittest/options/tst.aggrate-slow.r | 2 +- 4 files changed, 6 insertions(+), 6 deletions(-)
test/unittest/options/tst.aggrate-fast.d +2 -2
@@ -1,6 +1,6 @@ /* * Oracle Linux DTrace. - * Copyright (c) 2023, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2023, 2024, Oracle and/or its affiliates. All rights reserved. * Licensed under the Universal Permissive License v 1.0 as shown at * http://oss.oracle.com/licenses/upl. */
@@ -11,7 +11,7 @@ */ /* @@trigger: periodic_output */ -/* @@sort */ +/* @@nosort */ #pragma D option quiet #pragma D option switchrate=100ms
test/unittest/options/tst.aggrate-fast.r +1 -1
@@ -1,3 +1,4 @@ + 1 2
@@ -14,4 +15,3 @@ 8 -
test/unittest/options/tst.aggrate-slow.d +2 -2
@@ -1,6 +1,6 @@ /* * Oracle Linux DTrace. - * Copyright (c) 2023, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2023, 2024, Oracle and/or its affiliates. All rights reserved. * Licensed under the Universal Permissive License v 1.0 as shown at * http://oss.oracle.com/licenses/upl. */
@@ -10,7 +10,7 @@ * actions, multiple printa() should all reflect the same stale count. */ /* @@trigger: periodic_output */ -/* @@sort */ +/* @@nosort */ #pragma D option quiet #pragma D option switchrate=100ms
test/unittest/options/tst.aggrate-slow.r +1 -1
@@ -1,3 +1,4 @@ + 1 1
@@ -14,4 +15,3 @@ 5 -
Signature · 1 line
2.18.4
Re: [DTrace-devel] [PATCH 3/8] test: Fix nonexistent @@sort option
Quoted 3 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
Signed-off-by Eugene Loh <[email protected]>
Quoted 85 lines
--- test/unittest/options/tst.aggrate-fast.d | 4 ++-- test/unittest/options/tst.aggrate-fast.r | 2 +- test/unittest/options/tst.aggrate-slow.d | 4 ++-- test/unittest/options/tst.aggrate-slow.r | 2 +- 4 files changed, 6 insertions(+), 6 deletions(-)test/unittest/options/tst.aggrate-fast.d +2
-2@@ -1,6 +1,6 @@ /* * Oracle Linux DTrace. - * Copyright (c) 2023, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2023, 2024, Oracle and/or its affiliates. All rights reserved. * Licensed under the Universal Permissive License v 1.0 as shown at * http://oss.oracle.com/licenses/upl. */@@ -11,7 +11,7 @@ */ /* @@trigger: periodic_output */ -/* @@sort */ +/* @@nosort */ #pragma D option quiet #pragma D option switchrate=100mstest/unittest/options/tst.aggrate-fast.r +1
-1@@ -1,3 +1,4 @@ + 1 2@@ -14,4 +15,3 @@ 8 -test/unittest/options/tst.aggrate-slow.d +2
-2@@ -1,6 +1,6 @@ /* * Oracle Linux DTrace. - * Copyright (c) 2023, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2023, 2024, Oracle and/or its affiliates. All rights reserved. * Licensed under the Universal Permissive License v 1.0 as shown at * http://oss.oracle.com/licenses/upl. */@@ -10,7 +10,7 @@ * actions, multiple printa() should all reflect the same stale count. */ /* @@trigger: periodic_output */ -/* @@sort */ +/* @@nosort */ #pragma D option quiet #pragma D option switchrate=100mstest/unittest/options/tst.aggrate-slow.r +1
-1@@ -1,3 +1,4 @@ + 1 1@@ -14,4 +15,3 @@ 5 -Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
PATCH 4/8 test: Remove unneeded -w option
In reply to [email protected]From: Eugene Loh <[email protected]>
--- test/unittest/usdt/tst.dlclose2.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
test/unittest/usdt/tst.dlclose2.sh +1 -1
@@ -128,7 +128,7 @@ if [ $? -ne 0 ]; then fi script() { - $dtrace -w -c ./main -Zqs /dev/stdin <<EOF + $dtrace -c ./main -Zqs /dev/stdin <<EOF test_prov*::: { printf("%s:%s:%s\n", probemod, probefunc, probename);
Signature · 1 line
2.18.4
Re: [DTrace-devel] [PATCH 4/8] test: Remove unneeded -w option
Quoted 3 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
Signed-off-by Eugene Loh <[email protected]>
Quoted 25 lines
--- test/unittest/usdt/tst.dlclose2.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)test/unittest/usdt/tst.dlclose2.sh +1
-1@@ -128,7 +128,7 @@ if [ $? -ne 0 ]; then fi script() { - $dtrace -w -c ./main -Zqs /dev/stdin <<EOF + $dtrace -c ./main -Zqs /dev/stdin <<EOF test_prov*::: { printf("%s:%s:%s\n", probemod, probefunc, probename);Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
PATCH 5/8 Fix stddev() carryover computation
In reply to [email protected]From: Eugene Loh <[email protected]>
The stddev() aggregation function squares 64-bit data values. A value is split into 32-bit high and low values. Then, (high + low) is squared to produce high*high, 2*high*low, and low*low. Each is managed in its own 64-bit register, with the final result residing in two 64-bit registers.
When the 2*high*low portion is combined with the low*low portion, care is exercised in case the combination has a carryover portion to the higher bits. This check was broken in the case where low==0. That is, data values whose lowest 32 bits were 0 resulted in outrageously bad stddev() results.
Fix the check and add a test for such cases.
--- libdtrace/dt_cg.c | 2 +- test/unittest/aggs/tst.stddev2.d | 45 ++++++++++++++++++++++++++++++++ test/unittest/aggs/tst.stddev2.r | 13 +++++++++ 3 files changed, 59 insertions(+), 1 deletion(-) create mode 100644 test/unittest/aggs/tst.stddev2.d create mode 100644 test/unittest/aggs/tst.stddev2.r
libdtrace/dt_cg.c +1 -1
@@ -8290,7 +8290,7 @@ dt_cg_agg_stddev(dt_pcb_t *pcb, dt_ident_t *aid, dt_node_t *dnp,/* Add low value part from mid to lowreg */ emit (dlp,BPF_ALU64_REG (BPF_ADD , lowreg, lmdreg));/* Handle the overflow/carry case */ -emit (dlp,BPF_BRANCH_REG (BPF_JLT , lmdreg, lowreg, Lncy)); +emit (dlp,BPF_BRANCH_REG (BPF_JLE , lmdreg, lowreg, Lncy));emit (dlp,BPF_ALU64_IMM (BPF_ADD , hi_reg,1 ))/* account for carry */ ;/* Sum high value; no overflow expected nor accounted for */
test/unittest/aggs/tst.stddev2.d new file +45 -0
45 lines not on this page · raw message
test/unittest/aggs/tst.stddev2.r new file +13 -0
@@ -0,0 +1,13 @@ + + 0 + 8 + 80 + 800 + 8000 + 80000 + 800000 + 8000000 + 10000000 + 20000000 + 40000000 + 80000000
Signature · 1 line
2.18.4
Re: [DTrace-devel] [PATCH 5/8] Fix stddev() carryover computation
Quoted 14 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
The
stddev()aggregation function squares 64-bit data values. A value is split into 32-bit high and low values. Then, (high + low) is squared to produce high*high, 2*high*low, and low*low. Each is managed in its own 64-bit register, with the final result residing in two 64-bit registers.When the 2*high*low portion is combined with the low*low portion, care is exercised in case the combination has a carryover portion to the higher bits. This check was broken in the case where low==0. That is, data values whose lowest 32 bits were 0 resulted in outrageously bad
stddev()results.
Quoted 3 lines
Fix the check and add a test for such cases.
Signed-off-by Eugene Loh <[email protected]>
Quoted 99 lines
--- libdtrace/dt_cg.c | 2 +- test/unittest/aggs/tst.stddev2.d | 45 ++++++++++++++++++++++++++++++++ test/unittest/aggs/tst.stddev2.r | 13 +++++++++ 3 files changed, 59 insertions(+), 1 deletion(-) create mode 100644 test/unittest/aggs/tst.stddev2.d create mode 100644 test/unittest/aggs/tst.stddev2.rlibdtrace/dt_cg.c +1
-1@@ -8290,7 +8290,7 @@ dt_cg_agg_stddev(dt_pcb_t *pcb, dt_ident_t *aid, dt_node_t *dnp,/* Add low value part from mid to lowreg */ emit (dlp,BPF_ALU64_REG (BPF_ADD , lowreg, lmdreg));/* Handle the overflow/carry case */ -emit (dlp,BPF_BRANCH_REG (BPF_JLT , lmdreg, lowreg, Lncy)); +emit (dlp,BPF_BRANCH_REG (BPF_JLE , lmdreg, lowreg, Lncy));emit (dlp,BPF_ALU64_IMM (BPF_ADD , hi_reg,1 ))/* account for carry */ ;/* Sum high value; no overflow expected nor accounted for */ test/unittest/aggs/tst.stddev2.d new file +45
-0@@ -0,0 +1,45 @@ +/* + * Oracle Linux DTrace. + * Copyright (c) 2024, Oracle and/or its affiliates. All rights reserved. + * Licensed under the Universal Permissive License v 1.0 as shown at + * http://oss.oracle.com/licenses/upl. + */ + +/* + * ASSERTION: Positive stddev() test + * + * SECTION: Aggregations/Aggregations + * + * NOTES: This is a simple verifiable positive test of the stddev() function. + */ + +#pragma D option quiet + +BEGIN +{ + @a = stddev( 0); @a = stddev( 0); + @b = stddev( 0x10); @b = stddev( 0x20); + @c = stddev( 0x100); @c = stddev( 0x200); + @d = stddev( 0x1000); @d = stddev( 0x2000); + @e = stddev( 0x10000); @e = stddev( 0x20000); + @f = stddev( 0x100000); @f = stddev( 0x200000); + @g = stddev( 0x1000000); @g = stddev( 0x2000000); + @h = stddev( 0x10000000); @h = stddev( 0x20000000); + @i = stddev( 0x20000000); @i = stddev( 0x40000000); + @j = stddev( 0x40000000); @j = stddev( 0x80000000); + @k = stddev( 0x80000000); @k = stddev(0x100000000); + @l = stddev(0x100000000); @l = stddev(0x200000000); + printa("%9@x\n", @a); + printa("%9@x\n", @b); + printa("%9@x\n", @c); + printa("%9@x\n", @d); + printa("%9@x\n", @e); + printa("%9@x\n", @f); + printa("%9@x\n", @g); + printa("%9@x\n", @h); + printa("%9@x\n", @i); + printa("%9@x\n", @j); + printa("%9@x\n", @k); + printa("%9@x\n", @l); + exit(0); +}test/unittest/aggs/tst.stddev2.r new file +13
-0@@ -0,0 +1,13 @@ + + 0 + 8 + 80 + 800 + 8000 + 80000 + 800000 + 8000000 + 10000000 + 20000000 + 40000000 + 80000000Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
PATCH 6/8 test: Check dtrace return status
In reply to [email protected]From: Eugene Loh <[email protected]>
This test was spuriously passing. DTrace failed to run -- many error messages in the log file indicated as much -- and yet the test passed.
Check the dtrace return status. Due to recent USDT changes, the test now passes (properly).
--- test/unittest/usdt/tst.forker.sh | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
test/unittest/usdt/tst.forker.sh +5 -1
@@ -1,7 +1,7 @@ #!/bin/bash # # Oracle Linux DTrace. -# Copyright (c) 2007, Oracle and/or its affiliates. All rights reserved. +# Copyright (c) 2007, 2024, Oracle and/or its affiliates. All rights reserved. # Licensed under the Universal Permissive License v 1.0 as shown at # http://oss.oracle.com/licenses/upl. #
@@ -32,6 +32,10 @@ while kill -0 $id >/dev/null 2>&1; do exit(0); } EOF + if [ $? -ne 0 ]; then + echo ERROR: DTrace failed to run + exit 1 + fi done exit 0
Signature · 1 line
2.18.4
PATCH 7/8 Clean up double semicolons
In reply to [email protected]From: Eugene Loh <[email protected]>
--- dtprobed/dtprobed.c | 2 +- libdtrace/dt_cg.c | 2 +- libdtrace/dt_probe.c | 4 ++-- libdtrace/dt_prov_fbt.c | 2 +- libdtrace/dt_prov_lockstat.c | 4 ++-- libdtrace/dt_prov_profile.c | 2 +- 6 files changed, 8 insertions(+), 8 deletions(-)
dtprobed/dtprobed.c +1 -1
@@ -516,7 +516,7 @@ helper_ioctl(fuse_req_t req, int cmd, void *arg, gen =dof_stash_remove (pid, (uintptr_t ) arg);fuse_reply_ioctl (req, gen,NULL ,0 );return ; -default : errmsg ="invalid ioctl" ;; +default : errmsg ="invalid ioctl" ;fuse_log (FUSE_LOG_WARNING ,"%i: dtprobed: %s %x\n" , pid, errmsg, cmd); goto fuse_err;
libdtrace/dt_cg.c +1 -1
@@ -6562,7 +6562,7 @@ dt_cg_subr_inet_ntop(dt_node_t *dnp, dt_irlist_t *dlp, dt_regset_t *drp) {dt_node_t *af = dnp->dn_args;dt_node_t *addr = af->dn_list; -dt_node_t *tnp, *cnp, *lnp, *rnp, *anp, *xnp;; +dt_node_t *tnp, *cnp, *lnp, *rnp, *anp, *xnp;dt_idsig_t *isp;dt_decl_t *ddp;
libdtrace/dt_probe.c +2 -2
@@ -1345,7 +1345,7 @@ dt_probe_add_clause(dtrace_hdl_t *dtp, dt_probe_t *prp, dt_ident_t *idp) {dt_probe_clause_t *pcp; - pcp =dt_zalloc (dtp,sizeof (dt_probe_clause_t));; + pcp =dt_zalloc (dtp,sizeof (dt_probe_clause_t));if (pcp ==NULL )return dt_set_errno (dtp,EDT_NOMEM );
@@ -1395,7 +1395,7 @@ dt_probe_add_dependent(dtrace_hdl_t *dtp, dt_probe_t *prp, dt_probe_t *dprp)return 0 ; } - pdp =dt_zalloc (dtp,sizeof (dt_probe_dependent_t));; + pdp =dt_zalloc (dtp,sizeof (dt_probe_dependent_t));if (pdp ==NULL )return dt_set_errno (dtp,EDT_NOMEM );
libdtrace/dt_prov_fbt.c +1 -1
@@ -367,7 +367,7 @@ static int kprobe_attach(dtrace_hdl_t *dtp, const dt_probe_t *prp, int bpf_fd) EVENTSFS, FBT_GROUP_DATA, prp->desc->fun) + 1; fn =dt_alloc (dtp, len);if (fn ==NULL ) -return -ENOENT ;; +return -ENOENT ;snprintf (fn, len,"%s" FBT_GROUP_FMT "/%s/format" ,EVENTSFS ,FBT_GROUP_DATA , prp->desc->fun);
libdtrace/dt_prov_lockstat.c +2 -2
@@ -288,7 +288,7 @@ static int trampoline(dt_pcb_t *pcb, uint_t exitlbl)return 1 ; }else {int kind =1 ;/* reader (default) */ -uint_t lbl_reset =dt_irlist_label (dlp);; +uint_t lbl_reset =dt_irlist_label (dlp);if (strstr (uprp->desc->fun,"_write_" ) !=NULL ) kind =0 ;/* writer */
@@ -337,7 +337,7 @@ static int trampoline(dt_pcb_t *pcb, uint_t exitlbl)copy_lockaddr_into_cpuinfo (dlp,0 );return 1 ; }else { -uint_t lbl_reset =dt_irlist_label (dlp);; +uint_t lbl_reset =dt_irlist_label (dlp);if (strstr (uprp->desc->fun,"_trylock" ) !=NULL ) {/* The return value (arg1) must be 1. */
libdtrace/dt_prov_profile.c +1 -1
@@ -261,7 +261,7 @@ static int attach(dtrace_hdl_t *dtp, const dt_probe_t *prp, int bpf_fd) {profile_probe_t *pp = prp->prv_data;struct perf_event_attr attr; -int i, nattach =0 ;; +int i, nattach =0 ;int cnt =FDS_CNT (pp->kind);memset (&attr,0 ,sizeof (attr));
Signature · 1 line
2.18.4
Re: [DTrace-devel] [PATCH 7/8] Clean up double semicolons
Quoted 3 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
Signed-off-by Eugene Loh <[email protected]>
Quoted 113 lines
--- dtprobed/dtprobed.c | 2 +- libdtrace/dt_cg.c | 2 +- libdtrace/dt_probe.c | 4 ++-- libdtrace/dt_prov_fbt.c | 2 +- libdtrace/dt_prov_lockstat.c | 4 ++-- libdtrace/dt_prov_profile.c | 2 +- 6 files changed, 8 insertions(+), 8 deletions(-)dtprobed/dtprobed.c +1
-1@@ -516,7 +516,7 @@ helper_ioctl(fuse_req_t req, int cmd, void *arg, gen =dof_stash_remove (pid, (uintptr_t ) arg);fuse_reply_ioctl (req, gen,NULL ,0 );return ; -default : errmsg ="invalid ioctl" ;; +default : errmsg ="invalid ioctl" ;fuse_log (FUSE_LOG_WARNING ,"%i: dtprobed: %s %x\n" , pid, errmsg, cmd); goto fuse_err;libdtrace/dt_cg.c +1
-1@@ -6562,7 +6562,7 @@ dt_cg_subr_inet_ntop(dt_node_t *dnp, dt_irlist_t *dlp, dt_regset_t *drp) {dt_node_t *af = dnp->dn_args;dt_node_t *addr = af->dn_list; -dt_node_t *tnp, *cnp, *lnp, *rnp, *anp, *xnp;; +dt_node_t *tnp, *cnp, *lnp, *rnp, *anp, *xnp;dt_idsig_t *isp;dt_decl_t *ddp;libdtrace/dt_probe.c +2
-2@@ -1345,7 +1345,7 @@ dt_probe_add_clause(dtrace_hdl_t *dtp, dt_probe_t *prp, dt_ident_t *idp) {dt_probe_clause_t *pcp; - pcp =dt_zalloc (dtp,sizeof (dt_probe_clause_t));; + pcp =dt_zalloc (dtp,sizeof (dt_probe_clause_t));if (pcp ==NULL )return dt_set_errno (dtp,EDT_NOMEM );@@ -1395,7 +1395,7 @@ dt_probe_add_dependent(dtrace_hdl_t *dtp, dt_probe_t *prp, dt_probe_t *dprp)return 0 ; } - pdp =dt_zalloc (dtp,sizeof (dt_probe_dependent_t));; + pdp =dt_zalloc (dtp,sizeof (dt_probe_dependent_t));if (pdp ==NULL )return dt_set_errno (dtp,EDT_NOMEM );libdtrace/dt_prov_fbt.c +1
-1@@ -367,7 +367,7 @@ static int kprobe_attach(dtrace_hdl_t *dtp, const dt_probe_t *prp, int bpf_fd) EVENTSFS, FBT_GROUP_DATA, prp->desc->fun) + 1; fn =dt_alloc (dtp, len);if (fn ==NULL ) -return -ENOENT ;; +return -ENOENT ;snprintf (fn, len,"%s" FBT_GROUP_FMT "/%s/format" ,EVENTSFS ,FBT_GROUP_DATA , prp->desc->fun);libdtrace/dt_prov_lockstat.c +2
-2@@ -288,7 +288,7 @@ static int trampoline(dt_pcb_t *pcb, uint_t exitlbl)return 1 ; }else {int kind =1 ;/* reader (default) */ -uint_t lbl_reset =dt_irlist_label (dlp);; +uint_t lbl_reset =dt_irlist_label (dlp);if (strstr (uprp->desc->fun,"_write_" ) !=NULL ) kind =0 ;/* writer */ @@ -337,7 +337,7 @@ static int trampoline(dt_pcb_t *pcb, uint_t exitlbl)copy_lockaddr_into_cpuinfo (dlp,0 );return 1 ; }else { -uint_t lbl_reset =dt_irlist_label (dlp);; +uint_t lbl_reset =dt_irlist_label (dlp);if (strstr (uprp->desc->fun,"_trylock" ) !=NULL ) {/* The return value (arg1) must be 1. */ libdtrace/dt_prov_profile.c +1
-1@@ -261,7 +261,7 @@ static int attach(dtrace_hdl_t *dtp, const dt_probe_t *prp, int bpf_fd) {profile_probe_t *pp = prp->prv_data;struct perf_event_attr attr; -int i, nattach =0 ;; +int i, nattach =0 ;int cnt =FDS_CNT (pp->kind);memset (&attr,0 ,sizeof (attr));Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
PATCH 8/8 Turn some leading spaces into tabs
In reply to [email protected]From: Eugene Loh <[email protected]>
--- libdtrace/dt_bpf.c | 2 +- libdtrace/dt_pid.c | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-)
libdtrace/dt_bpf.c +1 -1
@@ -1140,7 +1140,7 @@ dt_bpf_load_prog(dtrace_hdl_t *dtp, const dt_probe_t *prp,snprintf (msg,sizeof (msg),"BPF program load for '%s:%s:%s:%s' failed" , - pdp->prv, pdp->mod, pdp->fun, pdp->prb); + pdp->prv, pdp->mod, pdp->fun, pdp->prb);if (err ==EPERM )return dt_bpf_lockmem_error (dtp, msg); dt_bpf_error(dtp, "%s: %s\n", msg,
libdtrace/dt_pid.c +4 -4
@@ -905,7 +905,7 @@ dt_pid_create_usdt_probes(dtrace_hdl_t *dtp, dt_proc_t *dpr, dtrace_probedesc_t provider = (dof_parsed_t *) p;if (!validate_dof_record (path, provider,DIT_PROVIDER , dof_buf_size, seen_size)) - goto parse_err; + goto parse_err; prv = provider->provider.name;
@@ -915,16 +915,16 @@ dt_pid_create_usdt_probes(dtrace_hdl_t *dtp, dt_proc_t *dpr, dtrace_probedesc_t probe = (dof_parsed_t *) p;if (!validate_dof_record (path, probe,DIT_PROBE , dof_buf_size, seen_size)) - goto parse_err; + goto parse_err; - mod = probe->probe.name; + mod = probe->probe.name; fun = mod +strlen (mod) +1 ; prb = fun +strlen (fun) +1 ; p += probe->size; seen_size += probe->size; -/* +/* * Now the parsed DOF for this probe's tracepoints. */ for (size_t j =0 ; j < probe->probe.ntp; j++) {
Signature · 1 line
2.18.4
Re: [DTrace-devel] [PATCH 8/8] Turn some leading spaces into tabs
Quoted 3 lines · eugene.loh--- via DTrace-devel
From: Eugene Loh <[email protected]>
Signed-off-by Eugene Loh <[email protected]>
Quoted 59 lines
--- libdtrace/dt_bpf.c | 2 +- libdtrace/dt_pid.c | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-)libdtrace/dt_bpf.c +1
-1@@ -1140,7 +1140,7 @@ dt_bpf_load_prog(dtrace_hdl_t *dtp, const dt_probe_t *prp,snprintf (msg,sizeof (msg),"BPF program load for '%s:%s:%s:%s' failed" , - pdp->prv, pdp->mod, pdp->fun, pdp->prb); + pdp->prv, pdp->mod, pdp->fun, pdp->prb);if (err ==EPERM )return dt_bpf_lockmem_error (dtp, msg); dt_bpf_error(dtp, "%s: %s\n", msg,libdtrace/dt_pid.c +4
-4@@ -905,7 +905,7 @@ dt_pid_create_usdt_probes(dtrace_hdl_t *dtp, dt_proc_t *dpr, dtrace_probedesc_t provider = (dof_parsed_t *) p;if (!validate_dof_record (path, provider,DIT_PROVIDER , dof_buf_size, seen_size)) - goto parse_err; + goto parse_err; prv = provider->provider.name;@@ -915,16 +915,16 @@ dt_pid_create_usdt_probes(dtrace_hdl_t *dtp, dt_proc_t *dpr, dtrace_probedesc_t probe = (dof_parsed_t *) p;if (!validate_dof_record (path, probe,DIT_PROBE , dof_buf_size, seen_size)) - goto parse_err; + goto parse_err; - mod = probe->probe.name; + mod = probe->probe.name; fun = mod +strlen (mod) +1 ; prb = fun +strlen (fun) +1 ; p += probe->size; seen_size += probe->size; -/* +/* * Now the parsed DOF for this probe's tracepoints. */ for (size_t j =0 ; j < probe->probe.ntp; j++) {Signature · 7 lines
2.18.4 _______________________________________________ DTrace-devel mailing list [email protected] https://oss.oracle.com/mailman/listinfo/dtrace-devel
Quoted 11 lines · On Tue, Jun 04, 2024 at 02:00:01PM -0400, [email protected] wrote:
From: Eugene Loh <[email protected]>
The aggpercpu test uses profile-* to fire aggregations on every CPU. But whether for good (offline CPUs) or bad (unreliable probe) reasons, a CPU might not aggregate any values.
Allow this test to skip over such CPUs. That is, while a BPF agg map might have the correct default value for the aggregation function, it may not see any anticipated data. The quality of profile-* probes should be checked by a different test.
Quoted 5 lines
Also, when computing
avg()andstddev()in the awk check program, truncate results to ints more often to mimic the algorithms in DTrace for more robust checks.Signed-off-by Eugene Loh <[email protected]>
Quoted 104 lines
--- test/unittest/aggs/tst.aggpercpu.sh | 62 +++++++++++++++++++++-------- 1 file changed, 45 insertions(+), 17 deletions(-)test/unittest/aggs/tst.aggpercpu.sh +45
-17@@ -46,7 +46,8 @@ fi awk ' # The expected value for the aggregation is aggval. # The expected value on a CPU is (m * cpu + b). - function check(label, aggval, m, b) { + # The default value on a CPU that did not fire is defval. + function check(label, aggval, m, b, defval) { # Check the aggregation over all CPUs. getline; print "check:", $0;@@ -54,12 +55,38 @@ awk ' # Check the per-CPU values. for (i = 1; i <= ncpu; i++) { - getline; - print "check:", $0; - if (match($0, "^ \\[CPU ") != 1 || - strtonum($2) != cpu[i] || - strtonum($3) != m * cpu[i] + b) - printf("ERROR: %s, agg per cpu %d, line: %s\n", label, cpu[i], $0); + do { + getline; + print "check:", $0; + + # pass is +1 for pass, -1 for error, 0 for skip CPU + pass = 0; + + if (match($0, "^ \\[CPU ") != 1) { + print "ERROR: no CPU to read" + pass = -1; + } else if (strtonum($2) > cpu[i]) { + print "ERROR: skipped over expected CPU" + pass = -1; + } else if (strtonum($2) == cpu[i]) { + if (strtonum($3) != m * cpu[i] + b) { + print "ERROR: wrong value" + pass = -1; + } else { + # right value + pass = +1; + } + } else if ($3 != defval) { + print "ERROR: wrong default value" + pass = -1; + } else { + # skip over CPU that apparently did not fire + pass = 0; + } + + if (pass == -1) + printf("ERROR: %s, agg per cpu %d, line: %s\n", label, cpu[i], $0); + } while (pass == 0); } }@@ -99,13 +126,14 @@ awk ' { # First we finish computing our estimates for avg and stddev. # (The other results require no further action.) + # (We keep truncating to ints to mimic DTrace algorithms.) - xavg /= xcnt; + xavg /= xcnt; xavg = int(xavg); - xstm /= xcnt; - xstd /= xcnt; + xstm /= xcnt; xstm = int(xstm); + xstd /= xcnt; xstd = int(xstd); xstd -= xstm * xstm; - xstd = int(sqrt(xstd)); + xstd = sqrt(xstd); xstd = int(xstd); # Sort the cpus.@@ -113,12 +141,12 @@ awk ' # Now read the results and compare. - check("cnt", xcnt, 0, 1); - check("avg", xavg, 10, 3); - check("std", xstd, 0, 0); - check("min", xmin, 30, -10); - check("max", xmax, 40, -15); - check("sum", xsum, 50, 0); + check("cnt", xcnt, 0, 1, "0"); + check("avg", xavg, 10, 3, "0"); + check("std", xstd, 0, 0, "0"); + check("min", xmin, 30, -10, "9223372036854775807"); + check("max", xmax, 40, -15, "-9223372036854775808"); + check("sum", xsum, 50, 0, "0"); printf("done\n"); }Signature · 1 line
2.18.4