From: Abhishek Thakur (Abhishek.Thakur@synchronous.net)
Date: Sat Sep 22 2001 - 17:16:22 EDT
Hi Charles, Restarting on your thread, as now I too am finding this issue as a bottleneck. Trying to flood ping the NET+Lx board is causing the kernel to just hang when 'ps' is done. With decently high network load, 'ps' crases on us occassionaly. Have you made any progress on the same? I tried to go through ps code (hardly 100 line sof code) and could not find the cause of problem. The fact that the kernel freezes => that the fault lies - either in the /proc filesystem implementation - or the code handling ethernet i/f at low level (DMA etc.) I have been trying to port UCD-SNMP which uses /proc/net/xxx for information about the network related attributes - and the system freezes once in a few tries - even without 'ps' being used. The irritating part is absence of crash report. I tend to believe that either the 'DMA part of ethernet driver' or the '/proc filesystem implementation' is the cause of trouble. but then i don't have much knowledge on linux kernel - and it would be great if someone more knowledgeble was there to provide her/his inputs. BTW we are using NET+Lx on our own board - quite similar to developer board - with 2.0.38-1pre7arm kernel. regards Abhishek > -----Original Message----- > From: Charles Steaderman [mailto:charlies@poliac.com] > Sent: Wednesday, August 15, 2001 12:40 PM > To: uclinux-dev@uclinux.org > Subject: Re: [uClinux-dev] uCLinux stability on NET+ARM > > > If you really want to stress the kernel, flood ping the > uClinux box (ping -f -s > 3000 x.x.x.x). The type ps from the console shell on the > uClinux box. On our > NetSilicon development box (as well as our hardware) the > kernel will panic and > lock up after a few ps commands. Please try this and let me > know what you find. > We have discovered other lockup conditions in the NET+ARM > ethernet controller as > well as a DMA transfer problem in the NET+ARM CPU related to > the SWAP assembly > instruction. Still waiting for updates from NetSilicon > although I am not holding > my breath :-) > > - Charlie > > Quoting Giuseppe Colucciello <Jcolucciello@cfl.rr.com>: > > > Charles, > > > > I have been using Net-Lx for the past 3 months and we are running an > > http server (thttp) on it and seems to be quite stable and > reliable. I > > am > > currently porting UpNp (Universal Plug and Play) from the OpenSource > > SDK and finding few problems with the compiler > (arm-uclinux-g??), but I > > think > > they are specific to the way the macros are formulated in the UpNp > > code... > > Stable, I think is as stable as most of the other micros implemented > > (ported) > > out there,,, > > My 2 cents (for now) > > -G > > > > ----- Original Message ----- > > From: "Charles Steaderman" <charlies@poliac.com> > > To: <uclinux-dev@uclinux.org> > > Sent: Wednesday, August 15, 2001 10:55 AM > > Subject: [uClinux-dev] uCLinux stability on NET+ARM > > > > > > > Has there been any progress on the stability of uCLinux on the > > NetSilicon > > > NET+ARM CPU? I haven't seen anything posted for quite a > while since > > we > > posted > > > this problem to NetSilicon 6 months ago. Crashing the OS by flood > > pinging > > and > > > typing ps from the console doesn't give me a "warm and > fuzzy" feeling > > about the > > > robustness of the NET+ARM port. Any information (or > suggested places > > for > > me to > > > look) are welcomed. BTW, I do not know the ARM > architecture very well, > > but > > I am > > > willing to help solve this problem in any way that I can. > > > > > > -- > > > charlies@poliac.com > > > Charles Steaderman > > > VP Engineering > > > Poliac Research Corporation > > > (V) 952.707.6245 > > > (M) 612.242.6364 > > > This message resent by the uclinux-dev@uclinux.org list server > > http://www.uClinux.org/ > > > > > > > This message resent by the uclinux-dev@uclinux.org list server > > http://www.uClinux.org/ > > > > > > -- > charlies@poliac.com > Charles Steaderman > VP Engineering > Poliac Research Corporation > (V) 952.707.6245 > (M) 612.242.6364 > This message resent by the uclinux-dev@uclinux.org list > server http://www.uClinux.org/ > This message resent by the uclinux-dev@uclinux.org list server http://www.uClinux.org/
This archive was generated by hypermail 2.1.4 : Thu Sep 19 2002 - 13:20:13 EDT