[Bug tapsets/2295] New: support conditionally present syscalls

[Bug tapsets/2295] New: support conditionally present syscalls

[fche at redhat dot com]

Some syscalls (e.g., nfsservctl) are defined fully only if a certain kconfig
setting is on (in this case, CONFIG_NFSD).  For configurations without that,
only a weak symbol definition is included in the kernel, which
elfutils/systemtap fails to deem suitable.

We should handle this case better.  This could involve adding CONFIG_FOO
conditionals to the script preprocessor (and then adding those conditions to
syscalls*.stp), or letting the weak symbols resolve specially somehow (point to
some dummy address?  or the general "syscall not implemented" function?).

Talk amongst yourselves.

-- 
           Summary: support conditionally present syscalls
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: tapsets
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: fche at redhat dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[bibo dot mao at intel dot com]

------- Additional Comments From bibo dot mao at intel dot com  2006-02-08 03:36 -------
I personally prefer weak symbols to dummy addresses or discard the weak 
symbols. And it will fit for different kernel version, kernel configuration 
option and architecture. But it is difficult to differentiate weak symbols and 
error function name.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-06-01 21:24 -------
The proposed optional-probe syntax in bug #2645 should provide a nice solution
for this problem.  Those syscalls that aren't always present could be marked as
optional.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |2645


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-06-01 21:31 -------
The same problem exists when trying to track changes between kernel versions. 
We could try to use the preprocessor to check the kernel version to
conditionally define a syscall, or we could mark such syscalls with the
'optional' syntax.

For example, when I run buildok/syscall.stp on i686 RHEL4, I get the following
errors.  The unresolved functions are all absent from this kernel.


semantic error: no match for probe point
while: resolving probe point kernel.function("sys_inotify_add_watch")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_inotify_init")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_inotify_rm_watch")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_ioprio_get")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_ioprio_set")
semantic error: no match for probe point
while: resolving probe point kernel.function("do_sched_setscheduler")
semantic error: no match for probe point
while: resolving probe point kernel.function("do_sched_setscheduler")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_set_zone_reclaim")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_shmat")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_inotify_add_watch").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_inotify_init").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_inotify_rm_watch").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_ioprio_get").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_ioprio_set").return
semantic error: no match for probe point
while: resolving probe point kernel.function("do_sched_setscheduler").return
semantic error: no match for probe point
while: resolving probe point kernel.function("do_sched_setscheduler").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_set_zone_reclaim").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_shmat").return
Pass 2: analysis failed.  Try again with more '-v' (verbose) options.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[joshua dot i dot stone at intel dot com]

-- 
Bug 2295 depends on bug 2645, which changed state.

Bug 2645 Summary: Need a syntax to specify optional probes
http://sourceware.org/bugzilla/show_bug.cgi?id=2645

           What    |Old Value                   |New Value
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
             Status|ASSIGNED                    |WAITING
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED

http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[joshua dot i dot stone at intel dot com]

------- Additional Comments From joshua dot i dot stone at intel dot com  2006-06-02 19:25 -------
On x86_64 FC5, buildok/syscall.stp gives me this:

semantic error: no match for probe point
while: resolving probe point kernel.function("sys_ftruncate64")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_truncate64")
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_ftruncate64").return
semantic error: no match for probe point
while: resolving probe point kernel.function("sys_truncate64").return
Pass 2: analysis failed.  Try again with more '-v' (verbose) options.


With the resolution of bug #2645, missing or conditionally-present syscalls can
probably be marked with "?", so that syscall.* can still work in their absence.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[joshua dot i dot stone at intel dot com]

-- 
Bug 2295 depends on bug 2645, which changed state.

Bug 2645 Summary: Need a syntax to specify optional probes
http://sourceware.org/bugzilla/show_bug.cgi?id=2645

           What    |Old Value                   |New Value
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |

http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[fche at redhat dot com]

-- 
Bug 2295 depends on bug 2645, which changed state.

Bug 2645 Summary: Need a syntax to specify optional probes
http://sourceware.org/bugzilla/show_bug.cgi?id=2645

           What    |Old Value                   |New Value
----------------------------------------------------------------------------
             Status|REOPENED                    |WAITING
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED

http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[eugen at debian dot org]

------- Additional Comments From eugen at debian dot org  2007-05-14 18:22 -------
Created an attachment (id=1835)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=1835&action=view)
patch that fixes handling of conditional syscalls for linux 2.6.22-rc1

This patch adds question marks to definitions of all condintional syscalls
found in kernel 2.6.22-rc1 (file sys_ni.c)

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/2295] support conditionally present syscalls

[hunt at redhat dot com]

------- Additional Comments From hunt at redhat dot com  2007-09-19 20:31 -------
I thought all the conditional syscalls had been fixed by now, but I was
surprised to find a few that have been missed. Checked those in and verified
that all in the attached patch are fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://sourceware.org/bugzilla/show_bug.cgi?id=2295

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.