Opened 5 years ago

Closed 3 years ago

#15264 closed bug (fixed)

KERN: debug_server: Killing team 142 (/boot/system/servers/app_server)

Reported by: bbjimmy Owned by: nobody
Priority: high Milestone: Unscheduled
Component: System Version: R1/Development
Keywords: Cc:
Blocked By: Blocking: #13554, #15309
Platform: x86

Description

hrev53379 browsing the web ... news.google.com

It seems that poor memory handeling causes low memory contitions ... eventually the debug server kills the app server for some reason. Obviously the system is frozen and myst be rebooted to get anywhere from there.

Attachments (3)

syslog (369.8 KB ) - added by bbjimmy 5 years ago.
A syslog showing the issue.
syslog_killQupzilla.txt (48.4 KB ) - added by bbjimmy 5 years ago.
This time it just killed QupZilla and left the app_server running
out.txt (357.5 KB ) - added by bbjimmy 5 years ago.
output from Haiku-analysis.cpp

Download all attachments as: .zip

Change History (6)

by bbjimmy, 5 years ago

Attachment: syslog added

A syslog showing the issue.

by bbjimmy, 5 years ago

Attachment: syslog_killQupzilla.txt added

This time it just killed QupZilla and left the app_server running

by bbjimmy, 5 years ago

Attachment: out.txt added

output from Haiku-analysis.cpp

comment:1 by waddlesplash, 5 years ago

hrev53406 should help here, saving about half of the latter section of fragmentation (so, a savings of 300-400MB.) The first part of the file, which totals a wastage of about 500MB, is tracked upstream as https://github.com/mjansson/rpmalloc/issues/117

comment:2 by waddlesplash, 5 years ago

Blocking: 15309 added

comment:3 by waddlesplash, 3 years ago

Blocking: 13554 added
Resolution: fixed
Status: newclosed

rpmalloc has not been used for a while, so this is "fixed". The question of whether to switch allocators is tracked in #13554.

Note: See TracTickets for help on using tickets.