Re: Kernel Panic in SANDPOINT 5.99.64-GENERIC on DSM-G600
Frank Wille <[email protected]> Tue, 14 Feb 2012 11:07:18 +0100
| Newsgroups | gmane.os.netbsd.ports.sandpoint |
|---|---|
| Organization | Privates Internet Ostwestfalen/Lippe |
| Message-ID | <[email protected]> |
On Mon, 13 Feb 2012 19:44:31 +0100 Charles Brown <[email protected]> wrote: > I'm experiencing a kernel panic on my DSM-G600 rev.B2. > 1) Enable ftp server via inetd in /etc/inetd.conf > 2) Connect via ethernet the DSM-G600 to a PC (my linux machine). > 3) Create a 1GB file using: dd if=/dev/zero of=1giga bs=1024 > count=1000000 in the linux machine. > 4) Connect to DSM-G600 via ftp and try to send him 1 giga using: put > 1giga Few seconds e...: > > dsm-g600# trap: kernel read DSI trap @ 0xd1a6752e by 0x995d0 (DSISR > 0x40000000, err=14), lr 0x99954 > panic: trap > cpu0: Begin traceback... > 0x00664a00: at panic+0x4c > 0x00664a40: at trap+0x104 > 0x00664ad0: kernel DSI read trap @ 0xd1a6752e by m_xhalf+0x9c: The crash happened in bpf_filter.c, accessing a page in kmem-space. I hope this is not another problem caused by the kmem/malloc changes after 5.99.60. > srr1=0x9030 r1=0x664ba0 cr=0x40002048 xer=0 ctr=0x9982c > dsisr=0x40000000 saved LR(0xffffefef) is invalid.cpu0: End Hmm. Corrupted stack frame? For a test you can disable the bpf(4) device. Comment out the line pseudo-device bpfilter # packet filter in your config and recompile the kernel. > traceback... dumpsys: TBD > * Yes... to be done... :| > [...] > Checking for core dump... > savecore: no core dump <-------- It really should be done. AFAIK no PPC port can currently dump a system core. > I tried also ftp get to DSM-G600, same result (build of the 1giga file > using dd on DSM-G600 works fine, so I think that is not a disk issue). > This happens every time. Also at the same or similar position in the transfer? > Could someone try to reproduce and confirm this? I tried it from an VIA-Epia board to a Synology DS-101g+, running 5.99.59. Transfer rates from only 100KB/s. There must be a problem with the Epia's i82540EM 1000BASE-T Ethernet adapter. But no crash. Then I tried it from a Soekris (AMD-Elan based) box to a QNAP TS-101, running 5.99.48. Transfer rates up to 800KB/s. But the Soekris is slow (133MHz). No crash again. In both cases all network interfaces run at 100baseTX full-duplex. Did you use Gigabit-Ethernet? I can try again with my G600 and a recent kernel this evening. -- Frank Wille