ed driver broken in today's -CURRENT?
I've just built a new kernel, based on a cvsup at 2030 UTC on 6 May, and since then *some* Ethernet transactions don't work. I've checked that it's not just a dead card: the previous kernel works fine. I have the funny situation that I can send fine, and I can traceroute to the box, but I can't ping. NFS also objects strenuously: 18:36:47.508811 panic.lemis.com freebie.lemis.com: icmp: echo request 18:36:47.508946 freebie.lemis.com panic.lemis.com: icmp: echo reply 18:36:48.512540 panic.lemis.com freebie.lemis.com: icmp: echo request 18:36:48.512669 freebie.lemis.com panic.lemis.com: icmp: echo reply 18:36:49.522568 panic.lemis.com freebie.lemis.com: icmp: echo request 18:36:49.522691 freebie.lemis.com panic.lemis.com: icmp: echo reply 18:36:50.532531 panic.lemis.com freebie.lemis.com: icmp: echo request 18:36:50.532664 freebie.lemis.com panic.lemis.com: icmp: echo reply 18:36:51.542504 panic.lemis.com freebie.lemis.com: icmp: echo request 18:36:51.542636 freebie.lemis.com panic.lemis.com: icmp: echo reply The above all looks good, but panic doesn't see the reply. 18:36:58.933420 freebie.lemis.com panic.lemis.com: icmp: echo request 18:36:59.953879 freebie.lemis.com panic.lemis.com: icmp: echo request 18:37:00.973929 freebie.lemis.com panic.lemis.com: icmp: echo request 18:37:01.993996 freebie.lemis.com panic.lemis.com: icmp: echo request 18:37:03.014053 freebie.lemis.com panic.lemis.com: icmp: echo request 18:37:04.034119 freebie.lemis.com panic.lemis.com: icmp: echo request panic seems to see nothing. 18:37:07.857977 freebie.lemis.com.57058 panic.lemis.com.33435: udp 12 [ttl 1] 18:37:07.858318 panic.lemis.com freebie.lemis.com: icmp: panic.lemis.com udp port 33435 unreachable 18:37:07.860258 freebie.lemis.com.57058 panic.lemis.com.33436: udp 12 [ttl 1] 18:37:07.860584 panic.lemis.com freebie.lemis.com: icmp: panic.lemis.com udp port 33436 unreachable 18:37:07.861455 freebie.lemis.com.57058 panic.lemis.com.33437: udp 12 [ttl 1] 18:37:07.861767 panic.lemis.com freebie.lemis.com: icmp: panic.lemis.com udp port 33437 unreachable This works fine! With the same kernel. After rebooting a kernel of a few days ago, panic can ping just fine: 18:39:17.258545 arp who-has panic.lemis.com tell panic.lemis.com 18:39:22.034446 arp who-has freebie.lemis.com tell panic.lemis.com 18:39:22.034543 arp reply freebie.lemis.com is-at 0:80:ad:b7:c9:c7 18:39:22.034875 panic.lemis.com freebie.lemis.com: icmp: echo request 18:39:22.034994 freebie.lemis.com panic.lemis.com: icmp: echo reply 18:39:23.040705 panic.lemis.com freebie.lemis.com: icmp: echo request 18:39:23.040837 freebie.lemis.com panic.lemis.com: icmp: echo reply 18:39:24.050692 panic.lemis.com freebie.lemis.com: icmp: echo request 18:39:24.050815 freebie.lemis.com panic.lemis.com: icmp: echo reply This is a Compex PCI-based NE2000 lookalike. It's on irq 3 (alone). Who gets the pointy hat? Greg -- Finger [EMAIL PROTECTED] for PGP public key See complete headers for address and phone numbers To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
On Sun, May 07, 2000 at 06:44:39PM +0930, Greg Lehey wrote: I've just built a new kernel, based on a cvsup at 2030 UTC on 6 May, and since then *some* Ethernet transactions don't work. I've checked that it's not just a dead card: the previous kernel works fine. I have the funny situation that I can send fine, and I can traceroute to the box, but I can't ping. NFS also objects strenuously: It is not dead card, it is broken TCP, see my similar report in -current, I notice it several hours ago right after TCP changes was commited. -- Andrey A. Chernov [EMAIL PROTECTED] http://nagual.pp.ru/~ache/ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
Eines schoenen Tages schrieb Andrey A. Chernov: On Sun, May 07, 2000 at 06:44:39PM +0930, Greg Lehey wrote: I've just built a new kernel, based on a cvsup at 2030 UTC on 6 May, and since then *some* Ethernet transactions don't work. I've checked that it's not just a dead card: the previous kernel works fine. I have the funny situation that I can send fine, and I can traceroute to the box, but I can't ping. NFS also objects strenuously: It is not dead card, it is broken TCP, see my similar report in -current, I notice it several hours ago right after TCP changes was commited. Same thing here, but the problems have gone to -stabe too. due to a probable tcp breakage i downgraded my machine to 4.0-stable which worked pretty good. with sources as from may 5th kernel boots fine, but tcp seems broken too. I experience a 50% package loss on out home lan. udp seems broken badly. rpc calls and nfs stopped working. i am using the vx driver on a 3com 3c597 EISA board. network card is ok, packet filtering rules have not been altered since the last kernel version. is nobody else seeing this? any clues? regards, oliver -- And remember: "To Infinity And Far Beyond ... Somehow?!" email: [EMAIL PROTECTED] [EMAIL PROTECTED] Hi! I'm a .signature virus! Copy me in your ~/.signature to help me spread! - Save this lifeform ;-) To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
Eines schoenen Tages schrieb Oliver Schonefeld: [snip] It is not dead card, it is broken TCP, see my similar report in -current, I notice it several hours ago right after TCP changes was commited. Same thing here, but the problems have gone to -stabe too. due to a probable tcp breakage i downgraded my machine to 4.0-stable which worked pretty good. with sources as from may 5th kernel boots fine, but tcp seems broken too. I experience a 50% package loss on out home lan. udp seems broken badly. rpc calls and nfs stopped working. i am using the vx driver on a 3com 3c597 EISA board. network card is ok, packet filtering rules have not been altered since the last kernel version. is nobody else seeing this? any clues? kind of looks like a problem in the delayed checksum calculation. while having a cvsup running and doing some nfs testing with a linux box pinging killed the -stable machine. (no flood pinging, just a normal ping) also nfs to the linux box was just beyond beeing awefully slow ... not even creeping is the reight term. kernel output from crash: delayed m_pullup, m-len: 84 off: 61420 p: 1 Fatal trap 12: page fault while in kernel mode fault virtual address = 0x8 fault code = supervisor read, page not present instruction pointer = 0x8:0xc0197f14 stack pointer = 0x10:0xc8934e34 frame pointer = 0x10:0xc8934e60 code segemnt= base 0x0, limit 0xf, type 0x1b = DPL 0, pres 1, def32 1, gran 1 processor eflags= interrupt enabled, resume, IOPL = 0 current procress= 213 (dnetc) interrupt mask = kernel: type 12, code=0 db trace ip_output(c0661e00,c8934ef4,c0661e00,c0661e4a,7) at ip_output+0xba0 ip_output(c0661e00,0,c8934ef4,14,0,0) at ip_output+0x5fb icmp_input(c0661e00,0) at icmp_input+0x716 icmp_input(c0661e00,c0661e00,fbdd809a,,40) at icmp_input+0x697 icmp_input(c0661e00,14,1,c0661e00,fbdd809a) at icmp_input+0x357 ip_input(c0661e00) at ip_input+0x780 ip_input(c01d7beb,0,2f,2f,2f) at ip_input+0x7df crashdump availabale on request. regards, oliver -- And remember: "To Infinity And Far Beyond ... Somehow?!" email: [EMAIL PROTECTED] [EMAIL PROTECTED] Hi! I'm a .signature virus! Copy me in your ~/.signature to help me spread! - Save this lifeform ;-) To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
-On [2507 14:50], Andrey A. Chernov ([EMAIL PROTECTED]) wrote: It is not dead card, it is broken TCP, see my similar report in -current, I notice it several hours ago right after TCP changes was commited. I assume you mean the NewReno commit submitted by Jayanth? -- Jeroen Ruigrok van der Werven Network- and systemadministrator [EMAIL PROTECTED]VIA Net.Works The Netherlands BSD: Technical excellence at its best http://www.via-net-works.nl And we are drunk with Death... To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
On Sun, May 07, 2000 at 04:15:57PM +0200, Jeroen Ruigrok van der Werven wrote: -On [2507 14:50], Andrey A. Chernov ([EMAIL PROTECTED]) wrote: It is not dead card, it is broken TCP, see my similar report in -current, I notice it several hours ago right after TCP changes was commited. I assume you mean the NewReno commit submitted by Jayanth? It is either NewReno or checksum changes, I not know which one. -- Andrey A. Chernov [EMAIL PROTECTED] http://nagual.pp.ru/~ache/ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
On Sun, May 07, 2000 at 04:15:57PM +0200, Jeroen Ruigrok van der Werven wrote: -On [2507 14:50], Andrey A. Chernov ([EMAIL PROTECTED]) wrote: It is not dead card, it is broken TCP, see my similar report in -current, I notice it several hours ago right after TCP changes was commited. I assume you mean the NewReno commit submitted by Jayanth? It is either NewReno or checksum changes, I not know which one. Since there's a sysctl you can use to turn the former off, perhaps it would have been smart to take a few seconds to narrow it down? -- \\ Give a man a fish, and you feed him for a day. \\ Mike Smith \\ Tell him he should learn how to fish himself, \\ [EMAIL PROTECTED] \\ and he'll hate you for a lifetime. \\ [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: ed driver broken in today's -CURRENT?
On Sun, 7 May 2000, Mike Smith wrote: Since there's a sysctl you can use to turn the former off, perhaps it would have been smart to take a few seconds to narrow it down? Those changes wouldn't have affected ICMP, but we tried that anyway. The problem was that the code changed the expression ~sum 0x to sum == 0x ? sum : ~sum 0x. I had just found and fixed this locally when I noticed Paul Saab committed the same functional fix. Well, it's nice to know that we can get multiple people finding the problem as soon as the symptoms of the breakage are known :) -- \\ Give a man a fish, and you feed him for a day. \\ Mike Smith \\ Tell him he should learn how to fish himself, \\ [EMAIL PROTECTED] \\ and he'll hate you for a lifetime. \\ [EMAIL PROTECTED] -- Brian Fundakowski Feldman \ FreeBSD: The Power to Serve! / [EMAIL PROTECTED]`--' To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message