OpenSolaris

Printable Version Enter a New Search
Bug ID 6588379
Synopsis brand interpose handlers stick around if a branded process dumps core
State 10-Fix Delivered (Fix available in build)
Category:Subcategory kernel:brandz
Keywords
Responsible Engineer Surya Prakki
Reported Against snv_66
Duplicate Of
Introduced In solaris_nevada
Commit to Fix snv_72
Fixed In snv_72
Release Fixed solaris_nevada(snv_72)
Related Bugs
Submit Date 2-August-2007
Last Update Date 31-August-2007
Description
If a branded process dumps core, the interpose handlers it installed tend to
stick around.
I reproduced the problem this way :
On a 4-way system [v40z], I bound the 'sleep' proces launched from an lx zone
to cpu2 and then made the sleep core dump [using QUIT].
Run this D cmd:

#dtrace -n brand_interpositioning_disable:entry'/cpu == 2/{trace(execname);}'

Until the above 1-liner prints out something, the IDT of cpu:2 will have the
handlers set this way :

128: brand_sys_int80                e030  3  + int  0
145: brand_sys_syscall_int          e030  3  + int  0

Interestingly, a kmdb breakpoint on brand_sys_syscall also gets a hit, in this window.
Work Around
N/A
Comments
N/A