[Bug 1001629] New: bsd stack uses wrong timeout values if hz != 100

[Bug 1001629] New: bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

           Summary: bsd stack uses wrong timeout values if hz != 100
           Product: eCos
           Version: CVS
          Platform: All
        OS/Version: All
            Status: UNCONFIRMED
          Severity: normal
          Priority: normal
         Component: TCP/IP
        AssignedTo: unassigned@bugs.ecos.sourceware.org
        ReportedBy: bernd.edlinger@hotmail.de
                CC: ecos-bugs@ecos.sourceware.org
             Class: Advice Request


Created an attachment (id=1842)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=1842)
proposed patch to solve this defect

if the eCos kernel is configured to 1000 IRQ/s all timeouts in the bsd_tcpip
stack expire in 1/10th of the expected time.

For instance all timeouts given with setsockopt(SOL_SOCKET, SO_RCVTIMEO),
which has a potential to break existing TCP/IP applications.

The reason is the wrong assumption that hz is always 100 in the file
"ecos/packages/net/bsd_tcpip/current/src/ecos/support.c".

Attached is a proposed patch to fix this issue
and a similar problem in show_ticks_in_us(),
where it is taken for granted, that the NOMINATOR in
CYGNUM_KERNEL_COUNTERS_RTC_RESOLUTION is always 1000000000 (1E9)
which is also not true in general.

-- 
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug 1001629] bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

Bernd Edlinger <bernd.edlinger@hotmail.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
               Flag|                            |Patch_or_Contribution+

-- 
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug 1001629] bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email, use the link below.

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

Grant Edwards <grant.b.edwards@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |grant.b.edwards@gmail.com

--- Comment #1 from Grant Edwards <grant.b.edwards@gmail.com> ---
I just ran into this exact same problem (though I noticed it when IPv6
addresses were timing out at 10% of the DHCP lease time).

The fix I'm currently using is almost identical to the proposed patch, except I
neglected to fix the show_ticks_in_us() function.

Is there some reason this fix hasn't been comitted?

What would it take to get it committed?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 1001629] bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email, use the link below.

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

Grant Edwards <grant.b.edwards@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Attachment #1842|0                           |1
        is obsolete|                            |

--- Comment #2 from Grant Edwards <grant.b.edwards@gmail.com> ---
Created attachment 2395
  --> http://bugs.ecos.sourceware.org/attachment.cgi?id=2395&action=edit
Add changelog entry to previous patch.

Same as previous patch, but with a Changelog entry added.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 1001629] bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email, use the link below.

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

Grant Edwards <grant.b.edwards@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned@bugs.ecos.source |grant.b.edwards@gmail.com
                   |ware.org                    |
     Ever confirmed|0                           |1

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 1001629] bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email, use the link below.

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

--- Comment #3 from Grant Edwards <grant.b.edwards@gmail.com> ---

I'd like to commit this patch.

I independently discovered this problem when the IPv6 stack was
prematurely timing out interface address assignments that had
non-infinite lifetimes.  I traced the problem to the exact same spot
and arrived at a nearly identical solution before I found this
bug/patch in Bugzilla.

I've tested the patch with both 10ms and 1ms system ticks, and I'm
satisfied that it's good (it's going into my production code).

Comments?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 1001629] New: bsd stack uses wrong timeout values if hz != 100

[bugzilla-daemon]

Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001629

           Summary: bsd stack uses wrong timeout values if hz != 100
           Product: eCos
           Version: CVS
          Platform: All
        OS/Version: All
            Status: UNCONFIRMED
          Severity: normal
          Priority: normal
         Component: TCP/IP
        AssignedTo: unassigned@bugs.ecos.sourceware.org
        ReportedBy: bernd.edlinger@hotmail.de
                CC: ecos-bugs@ecos.sourceware.org
             Class: Advice Request


Created an attachment (id=1842)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=1842)
proposed patch to solve this defect

if the eCos kernel is configured to 1000 IRQ/s all timeouts in the bsd_tcpip
stack expire in 1/10th of the expected time.

For instance all timeouts given with setsockopt(SOL_SOCKET, SO_RCVTIMEO),
which has a potential to break existing TCP/IP applications.

The reason is the wrong assumption that hz is always 100 in the file
"ecos/packages/net/bsd_tcpip/current/src/ecos/support.c".

Attached is a proposed patch to fix this issue
and a similar problem in show_ticks_in_us(),
where it is taken for granted, that the NOMINATOR in
CYGNUM_KERNEL_COUNTERS_RTC_RESOLUTION is always 1000000000 (1E9)
which is also not true in general.

-- 
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.