Bug #362: 08_1 model status: module_put issues - C Instrumentation Framework - Open-Source Projects

Actions

Copy link

Bug #362

closed

08_1 model status: module_put issues

Added by Evgeny Novikov over 13 years ago. Updated over 11 years ago.

Status:

Closed

Priority:

Normal

Assignee:

Evgeny Novikov

Category:

Start date:

08/05/2010

Due date:

% Done:

Estimated time:

Detected in build:

old bug

Platform:

Published in build:

e2a8d1f

Description

Model is implemented with some lacks (see bug #347).

Related issues 3 (0 open — 3 closed)

Actions

Copy link

Updated by Vadim Mutilin over 13 years ago

Aspectator does not instrument calls of module_put. It is interpreted as undefined function body.
See pppox.c

http://shved:8999/stats/errortrace/profilename/git/name/ldv_git/user/pavel/host/localhost/password/no/page/Error+trace/value/15943/Task+name/Kernel-59365d136d205cc20fe666ca7f89b1c5001b0d5a/Environment+version/artificial_env_1/Rule+name/08_1/Module/pppox.ko/Entry+point/ldv_main0_plain_sorted_withcheck

Actions

Copy link

Updated by Evgeny Novikov over 13 years ago

Priority changed from Low to High

This should be fixed in near future. Possibly on the next week.

Actions

Copy link

Updated by Evgeny Novikov over 13 years ago

In my mind the problem isn't with aspectator, it's connected with kernel core API changes as was for spin_lock. So the model should be fixed. But one day investigation and fixing is required.

Actions

Copy link

Updated by Alexey Khoroshilov about 13 years ago

Subject changed from 08_1 model status to 08_1 model status: module_put issues

I see the same issue with cpufreq.ko under ldv-git.

Actions

Copy link

Updated by Evgeny Novikov about 13 years ago

With environment variable LDV_PRINT_SIGNATURE_OF_MATCHED_BY_NAME=1 specified I obtain following warnings:

These functions were matched by name but have different signatures: 
source function declaration: extern void module_put (struct module *) 
aspect function declaration: void module_put (..)

As was mentioned it's related with #347 issue.

Actions

Copy link

Updated by Evgeny Novikov about 13 years ago

Published in build set to master

Works for master. See comments on #347 and #786 issues!

Actions

Copy link

Updated by Pavel Shved about 13 years ago

This fix made our nightly middle tests fail. Yesterday traces showed that module_put was "unimplemented", and today it's implemented! Two (flase) unsafes became safes:

drivers/kbdrivers/0008-2.6.31.6/test-0008-2.6.31.6-verdict-unsafe-drivers--net--pppox.c/pppox.ko
drivers/kbdrivers/0008-2.6.31.6/test-0008-2.6.31.6-verdict-unsafe-drivers--pci--hotplug--acpiphp_core.c/acpiphp_core.ko

See comparison (will be gone in 30 days):
http://10.10.2.59:8999/stats/comparison/task+ids/98+99/name/ldv_qa_night_res/user/qarobot/host/10.10.2.59/password/12345

Please, update the tests!

Actions

Copy link

Updated by Evgeny Novikov about 13 years ago

Priority changed from High to Normal

Yes, model becomes more correct, but I still see that it fails. For example, for driver drivers/kbdrivers/0008-2.6.31.6/test-0008-2.6.31.6-verdict-unsafe-drivers--mtd--mtd_blkdevs.c/mtd_blkdevs.ko (it's the first driver in the list of verified by night tester) module_put isn't instrumented. I try it locally but even LDV_PRINT_SIGNATURE_OF_MATCHED_BY_NAME says nothing. It's most likely the another problem in aspectator but it'll be addressed later...

Actions

Copy link

Updated by Evgeny Novikov almost 13 years ago

Status changed from Open to Resolved
Published in build changed from master to e2a8d1f

There is no more the given issue since we have moved aspectator from LLVM to GCC.

As I understand, the problem was related with incorrect walking through GCC internal representation in instrumenting. At the moment this is fixed and the fact that this model works good shows it.

Actions

Copy link

#10