Comments (4)
Thanks for the diagnosis, @JainTwinkle. I just looked and threadlist certainly has a couple of discrepancies around pid handling. I'll create a PR to clean it up.
from dmtcp.
Information from a GDB session:
(gdb) bt
#0 TLSInfo_GetTidOffset () at tls.cpp:248
#1 0x00007ffff7ecd149 in TLSInfo_GetPidOffset () at tls.cpp:300
#2 0x00007ffff7ecd234 in TLSInfo_VerifyPidTid (pid=8518, tid=8518) at tls.cpp:604
#3 0x00007ffff7ed2e68 in dmtcp::ThreadList::init () at threadlist.cpp:179
#4 0x00007ffff7ebe956 in dmtcp_initialize_entry_point () at dmtcpworker.cpp:293
#5 0x00007ffff7dea9c3 in call_init (env=0x7fffffffdd08, argv=0x7fffffffdcf8, argc=1, l=<optimized out>) at dl-init.c:82
#6 _dl_init (main_map=0x7ffff7ffe150, argc=1, argv=0x7fffffffdcf8, env=0x7fffffffdd08) at dl-init.c:131
#7 0x00007ffff7ddc17a in _dl_start_user () from /lib64/ld-linux-x86-64.so.2
#8 0x0000000000000001 in ?? ()
#9 0x00007fffffffdfd9 in ?? ()
#10 0x0000000000000000 in ?? ()
(gdb) l
243 * 'struct pthread' is zeroed out before adding tid and pid.
244 * pthread_desc below is defined as 'struct pthread' in glibc:nptl/descr.h
245 */
246 tmp = memsubarray((char *)pthread_desc, (char *)&tid_pid, sizeof(tid_pid));
247
248 if (tmp == NULL && glibcMajorVersion() == 2 && glibcMinorVersion() >= 24) {
249 // starting with glibc-2.25 (including 2.24.90 on Fedora), the pid field
250 // is deprecated and set to zero.
251 tid_pid.pid = 0;
252 tmp = memsubarray((char*)pthread_desc, (char *)&tid_pid, sizeof(tid_pid));
(gdb) l
253 }
254
255 if (tmp == NULL) {
256 JWARNING(false) (tid_pid.tid)
257 .Text("Couldn't find offsets of tid/pid in thread_area.\n"
258 " Now relying on the value determined using the\n"
259 " glibc version with which DMTCP was compiled.");
260 return STATIC_TLS_TID_OFFSET();
261
262 // mtcp_abort();
(gdb) p tmp
$3 = 0x0
(gdb) p glibcMinorVersion()
$6 = 17
(gdb) p glibcMajorVersion()
$7 = 2
from dmtcp.
Further debugging shows that pthread_desc
(array) and tid_pid
(subarray) do not match in memsubarray
function:
(gdb) l 319
314 // Assume subarray length is at least sizeof(int) and < 2048.
315 JASSERT(len >= sizeof(int));
316 for (i_ptr = array; i_ptr < array + 2048; i_ptr++) {
317 if (*(int *)i_ptr == word1) {
318 for (j = 0; j < len; j++) {
319 if (i_ptr[j] != subarray[j]) {
320 break;
321 }
322 }
323 if (j == len) {
(gdb) p j
$23 = 4
(gdb) p i_ptr[j]
$19 = 70 'F'
(gdb) p subarray[j]
$20 = 64 '@'
(gdb) x/8c subarray
0x7fffffffd698: 70 'F' 33 '!' 0 '\000' 0 '\000' 64 '@' -100 '\234' 0 '\000' 0 '\000'
(gdb) x/8c i_ptr
0x7ffff7e4aa50: 70 'F' 33 '!' 0 '\000' 0 '\000' 70 'F' 33 '!' 0 '\000' 0 '\000'
(gdb) p *(int *)i_ptr
$21 = 8518
(gdb) p *(int *)subarray
$22 = 8518
Hope this helps! Thanks!
from dmtcp.
I just realized that THREAD_REAL_PID
returns the virtual PID, not the real PID. And that's why the PID field doesn't match in the pthread_desc
that has the real_pid
after TID.
(gdb) p tid_pid
$32 = {tid = 8518, pid = 40000}
Relevant backtrace:
#0 getpid () at pid/pidwrappers.cpp:126
#1 0x00007ffff7e8590c in syscall (sys_num=<optimized out>) at pid/pid_miscwrappers.cpp:332
#2 0x00007ffff7f09238 in _real_syscall (sys_num=<optimized out>) at syscallsreal.c:891
#3 0x00007ffff7ed1990 in _real_getpid () at threadlist.cpp:96
#4 dmtcp_get_real_pid () at threadlist.cpp:96
#5 0x00007ffff7ecccd7 in TLSInfo_GetTidOffset () at tls.cpp:227
...
(gdb) p _dmtcp_pid
$37 = 40000
from dmtcp.
Related Issues (20)
- How to debug programs which reliably crash after snapshot restauration
- DMTCP checkpoint restore is slow
- port dmtcp for windows HOT 9
- Support for RISC-V HOT 3
- Create checkpoint from a dump file or a running process HOT 1
- `make check` fails on an ARMv8 machine HOT 1
- Segmentation fault at restart
- dmtcp in docker on apple silicon HOT 5
- dmtcp CI is broken for master branch - root cause: python3.8 vs python3.10 pty module HOT 2
- add soversion
- dmtcp build failed on ppc64le, aarch64 and s390x architecture
- The last few checkpoints of the dmtcp save are particularly slow
- INSTALL.md mentions non-existing command line option --no-coordinator for dmtcp_launch HOT 1
- Release 3.0, Windows Subsystem for Linux with Ubuntu 22.04 LTS: all checks fail HOT 1
- Release 3.0, Windows Subsystem for Linux with Ubuntu 22.04 LTS: restart doesn't work HOT 1
- DMTCP build is broken with recent PR 1061 HOT 1
- Duplicating(forking) a checkpointed process? HOT 2
- Segmentation fault on dmtcp (2.6.0) using MPICH (4.2.0)
- Segfault when I set 2 ckpts in a program using share memory HOT 1
- Using '--enable-logging' leads to hang in a simplest case HOT 5
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dmtcp.