Bug 738729 - stuck in endless loop in bwstat_delay when using spice client
Summary: stuck in endless loop in bwstat_delay when using spice client
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: trickle
Version: 16
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: Nicoleau Fabien
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-09-15 17:08 UTC by Alon Levy
Modified: 2014-08-04 22:08 UTC (History)
4 users (show)

Fixed In Version: trickle-1.07-12.fc16
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2011-10-09 19:51:38 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
bwstat_getdelay exit while loop if no packets (677 bytes, patch)
2011-09-19 08:05 UTC, Alon Levy
no flags Details | Diff

Description Alon Levy 2011-09-15 17:08:41 UTC
Description of problem:
The command:
 trickle -u 10 -d 20 `which spicec` -h localhost -p 9011

Becomes unresponsive. This doesn't happen if I raise the download limit enough (500+). Attaching with gdb shows being in an infinite loop inside a recv call on bwstat_delay. The attached patch to the package includes a simple patch to fix it. I've tested it with the above command and while it solves it, I get occational crashes so not sure if it is the right approach, but the crashes seem to be unrelated, and only happened when using gnome-shell (not sure how that relates, but possibly gnome-shell is causing a refresh of spice client that is causing it to try a recv - don't know).

Version-Release number of selected component (if applicable):
trickle-1.07-11.fc16.x86_64

How reproducible:
50% with the command line below (1/2), 100% with a win7 vm running qxl driver.

Steps to Reproduce:
1. start spice server (qemu-kvm -spice disable-ticketing,port=9011)
2. trickle -u 10 -d 20 `which spicec` -h localhost -p 9011
  
Actual results:
program gets stuck, no keyboard interaction possible. attaching shows stuck in 
/usr/src/debug/trickle-1.07/bwstat.c in the loop starting at line 189.

Expected results:
work normally (window appears with bios, can do Ctrl+B, press tab, play with the command line of PXE)

Additional info:

Following patch (to the fedora trickle package, includes the patch to trickle 1.07, tested on Fedora 16) fixes the problem:


commit 4bd0d811fadd7789295dbf5a4304f5159637584e
Author: Alon Levy <alevy>
Date:   Thu Sep 15 18:08:55 2011 +0300

    1.07-12 fix endless loop (local patch)

diff --git a/trickle-1.07-bwsta_getdelay-stop-if-no-packets.patch b/trickle-1.07-bwsta_getdelay-stop-if-no-packets.patch
new file mode 100644
index 0000000..6760350
--- /dev/null
+++ b/trickle-1.07-bwsta_getdelay-stop-if-no-packets.patch
@@ -0,0 +1,25 @@
+From 3b22c327ff6ac3ee51332919e91ae63d47225a9b Mon Sep 17 00:00:00 2001
+From: Alon Levy <alevy>
+Date: Thu, 15 Sep 2011 18:05:13 +0300
+Subject: [PATCH] bwsta_getdelay: stop if no packets
+
+---
+ bwstat.c |    2 +-
+ 1 files changed, 1 insertions(+), 1 deletions(-)
+
+diff --git a/bwstat.c b/bwstat.c
+index a1c1085..9567275 100644
+--- a/bwstat.c
++++ b/bwstat.c
+@@ -210,7 +210,7 @@ bwstat_getdelay(struct bwstat *bs, size_t *len, uint lim, short which)
+ 
+ 			ent += xent;
+ 		}
+-	} while (pool > 0 && ncli > 0);
++	} while (pool > 0 && ncli > 0 && (TAILQ_FIRST(&poolq) != TAILQ_END(&poolq)));
+ 
+ 	/*
+ 	 * This is the case of a client that is not using its limit.
+-- 
+1.7.6.2
+
diff --git a/trickle.spec b/trickle.spec
index 1e883aa..530b9f1 100644
--- a/trickle.spec
+++ b/trickle.spec
@@ -1,6 +1,6 @@
 Name:           trickle
 Version:        1.07 
-Release:        11%{?dist}
+Release:        12%{?dist}
 Summary:        Portable lightweight userspace bandwidth shaper
 
 Group:          Applications/System
@@ -15,6 +15,7 @@ BuildRequires:  libevent-devel
 Patch0:         %{name}-%{version}-include_netdb.patch
 Patch1:         %{name}-%{version}-libdir.patch
 Patch2:         %{name}-%{version}-CVE-2009-0415.patch
+Patch3:         %{name}-%{version}-bwsta_getdelay-stop-if-no-packets.patch
 
 %description
 trickle is a portable lightweight userspace bandwidth shaper.
@@ -33,6 +34,7 @@ trickle runs entirely in userspace and does not require root privileges.
 %patch0 -p1 -b .include_netdb
 %patch1 -p1 -b .libdir
 %patch2 -p1 -b .cve
+%patch3 -p1
 touch -r configure aclocal.m4 Makefile.in stamp-h.in
 
 iconv -f ISO88591 -t UTF8 < README > README.UTF8
@@ -71,6 +73,8 @@ rm -rf $RPM_BUILD_ROOT
 
 
 %changelog
+* Thu Sep 15 2011 Alon Levy <alevy> 1.07-12
+- Fix endless loop in bwstat_delay (seen with spicec)
 * Sun Feb 13 2011 Nicoleau Fabien <nicoleau.fabien> 1.07-11
 - Revert to the working patch
 * Wed Feb 09 2011 Fedora Release Engineering <rel-eng.org> - 1.07-10

Comment 1 Nicoleau Fabien 2011-09-18 21:02:34 UTC
Hello,
thank you for the report, and the patch.

But I don't really understand how to apply this. I tried to copy in a file, then applied it. But when I build, the patch doesn't work.

Can you simply provide the bwsta_getdelay-stop-if-no-packets.patch ?

Comment 2 Alon Levy 2011-09-19 08:05:06 UTC
Created attachment 523795 [details]
bwstat_getdelay exit while loop if no packets

Comment 3 Fedora Update System 2011-09-19 19:37:08 UTC
trickle-1.07-12.fc16 has been submitted as an update for Fedora 16.
https://admin.fedoraproject.org/updates/trickle-1.07-12.fc16

Comment 4 Fedora Update System 2011-09-20 19:03:50 UTC
Package trickle-1.07-12.fc16:
* should fix your issue,
* was pushed to the Fedora 16 testing repository,
* should be available at your local mirror within two days.
Update it with:
# su -c 'yum update --enablerepo=updates-testing trickle-1.07-12.fc16'
as soon as you are able to.
Please go to the following url:
https://admin.fedoraproject.org/updates/trickle-1.07-12.fc16
then log in and leave karma (feedback).

Comment 5 Fedora Update System 2011-10-09 19:51:38 UTC
trickle-1.07-12.fc16 has been pushed to the Fedora 16 stable repository.  If problems still persist, please make note of it in this bug report.


Note You need to log in before you can comment on or make changes to this bug.