Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

j9vmem__testOverlappingSegments in cmdLineTester_pltest_0 seems to segfault if unable to allocate any memory at all #19509

Open
cjjdespres opened this issue May 16, 2024 · 2 comments · May be fixed by #19517
Open

j9vmem__testOverlappingSegments in cmdLineTester_pltest_0 seems to segfault if unable to allocate any memory at all #19509

cjjdespres opened this issue May 16, 2024 · 2 comments · May be fixed by #19517

Comments

@cjjdespres
Copy link
Contributor

I have been running sanity.functional locally (in containers in a fyre VM, anyway) and have been getting a segfault in cmdLineTester_pltest_0. I believe this is because the subtest j9vmem__testOverlappingSegments is unable to allocate any memory at all:

 [ERR] Starting test j9vmem__testOverlappingSegments
 [ERR]   Cycles: 10000
 [ERR] 
 [ERR]   Segment           Start             End               Size              Keep cycle
 [ERR]   Failed to get memory. Error: Cannot allocate memory.
 [ERR]   Ignoring memory allocation failure(0 of 10000 loops finished).
 [ERR]   
 [ERR] =========================
 [ERR]   0 cycles completed
 [ERR]   0 segments freed
 [ERR] 
 [ERR] 
 [ERR] An unhandled error (24) has occurred.
 [ERR] J9Generic_Signal_Number=00000018
 [ERR] Signal_Number=0000000b
 [ERR] Error_Value=00000000
 [ERR] Signal_Code=00000001
 [ERR] Handler1=00005565C7619FE0
 [ERR] Handler2=00007FDFA7B00B40
 [ERR] InaccessibleAddress=0000000000000018
 [ERR] RDI=0000000000000018
 [ERR] RSI=0000000000000001
 [ERR] RAX=00007FDFA7B49740
 [ERR] RBX=0000000000000000
 [ERR] RCX=00007FDFA79ADA7B
 [ERR] RDX=0000000000000000
 [ERR] R8=0000000000000000
 [ERR] R9=0000000000000000
 [ERR] R10=0000000000000000
 [ERR] R11=0000000000000293
 [ERR] R12=0000000000001000
 [ERR] R13=00000000FFFFFFFF
 [ERR] R14=0000000000000000
 [ERR] R15=0000000000000000
 [ERR] RIP=00007FDFA7B27598
 [ERR] RSP=00007FFD81BE1568
 [ERR] EFlags=0000000000010246
 [ERR] RBP=0000000000000000
 [ERR] ERR=0000000000000004
 [ERR] TRAPNO=000000000000000E
 [ERR] OLDMASK=0000000000000000
 [ERR] CR2=0000000000000018
 [ERR] xmm0=0000000000000000
 [ERR] xmm1=0000000000000000
 [ERR] xmm2=0053534552444441
 [ERR] xmm3=415353454d5f434c
 [ERR] xmm4=657461636f6c6c61
 [ERR] xmm5=7272457473616c09
 [ERR] xmm6=7365747962203030
 [ERR] xmm7=6f7420656c62616e
 [ERR] xmm8=2e74657375706300
 [ERR] xmm9=0000000000000000
 [ERR] xmm10=0000001000000010
 [ERR] xmm11=00fe00fe00fe00fe
 [ERR] xmm12=00fe00fe00fe00fe
 [ERR] xmm13=0000000000000000
 [ERR] xmm14=0000000000000000
 [ERR] xmm15=0000000000000000
 [ERR] Module=/root/hostdir/openj9-openjdk-jdk17/build/linux-x86_64-server-release/images/jdk/bin/../lib/default/libj9prt29.so
 [ERR] Module_base_address=00007FDFA7AD7000

The backtrace I get is this:

#0  subtractAtomic (address=0x18, value=1) at /root/hostdir/openj9-openjdk-jdk17/omr/util/omrutil/AtomicFunctions.cpp:65
#1  0x00007fc2a17261e6 in omrmem_categories_decrement_counters (category=category@entry=0x0, size=size@entry=4096) at /root/hostdir/openj9-openjdk-jdk17/omr/port/common/omrmemcategories.c:97
#2  0x00007fc2a173faba in omrvmem_free_memory (portLibrary=<optimized out>, address=<optimized out>, byteAmount=4096, identifier=<optimized out>)
    at /root/hostdir/openj9-openjdk-jdk17/omr/port/linux/omrvmem.c:761
#3  0x0000564605d4ea61 in j9vmem_testOverlappingSegments ()
#4  0x0000564605d4ed0a in j9vmem_runTests ()
#5  0x0000564605d50178 in signalProtectedMain ()
#6  0x00007fc2a172d7f1 in omrsig_protect (portLibrary=0x7ffd3339be70, fn=0x564605d4f6c0 <signalProtectedMain>, fn_arg=0x7ffd3339be40, handler=0x564605d5afe0 <genericSignalHandler>, handler_arg=0x0, 
    flags=504, result=0x7ffd3339be20) at /root/hostdir/openj9-openjdk-jdk17/omr/port/unix/omrsignal.c:425
#7  0x0000564605d22a07 in main ()

(My debug information seems to be messed up a little - sorry).

I think the failure is happening here:

openj9/runtime/tests/port/j9vmemTest.c

Lines 3458 to 3468 in a56cbdf

/* Free remaining segments */
freed = 0;
for (j = 0; j < CYCLES; j++) {
if (keepCycles[j] >= i) {
I_32 rc = j9vmem_free_memory(vmemID[j].address, vmemParams[j].byteAmount, &vmemID[j]);
if (0 == rc) {
freed++;
}
}
}
outputComment(PORTLIB, "%d remaining segments freed\n\n", freed);

and what's happening is:

  1. We fail in the first cycle without ever allocating anything, then jump to exit (so i == 0).
  2. We have that keepCycles[0] >= 0 because keepCycles was memset to 0 earlier in the function
  3. We try to free memory that was never allocated, eventually crashing because the category of the memory is NULL.
@cjjdespres cjjdespres linked a pull request May 17, 2024 that will close this issue
Open
@cjjdespres
Copy link
Contributor Author

The problem was actually a little worse - if the test failed in the very first cycle, the cleanup procedure would attempt to deallocate every segment j from 1 to CYCLES - 1, because keepCycles[j] would always be zero. The same erroneous freeing would happen for segment 0 only if the initial memory allocation failed, like I said above.

@cjjdespres
Copy link
Contributor Author

cjjdespres commented May 17, 2024 •
edited

The test cmdLineTester_pltest_0 still fails in various other ways for me, presumably because I'm running in containers (or I've set up the tests incorrectly), but #19517 does fix this particular issue.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Avoid segfaults in j9vmem_testOverlappingSegments cjjdespres/openj9
1 participant
@cjjdespres