Alan Cox wrote:
> Boot Time Failures
> ------------------
> Use PCI DMA 'lost interrupt' problem with some hw [which ?]
> (NEC Versa LX with PIIX tuning)
Host bridge: Intel Corporation 440BX/ZX - 82443BX/ZX Host bridge (rev 3).
IDE interface: Intel Corporation 82371AB PIIX4 IDE (rev 1).
Sorry, can't do lspci, /proc doesn't have bus #2 in it and lspci aborts.
> Fix Exists But Isnt Merged
> --------------------------
> Loopback fs hangs
Is this fixed in ac16 or 17? My brother was working on mounted ISO images all last night
with no problems.
> To Do
> -----
> Tulip hang on rmmod/crashes sometimes
always on the second instance.
> PCMCIA/Cardbus hangs, IRQ problems, Keyboard/mouse problem (may be fixed ?)
Sorry :( not fixed but I have more information, see next item.
> To Do But Non Showstopper
> -------------------------
> PCMCIA crashes on unloading pci_socket
dhinds pcmcia stuff works incredibly better wrt to this.
> Compatibility Errors
> --------------------
> Xterm broke in 2.3.99pre6 (FIONREAD/select loop)
I found a bug in fwrite() I believe. fopen() a file, a+, seek to the beginning and write to
it. The write succeeds according to the return value. It should fail if the fpos pointer
is set before the end of the file. Should this goto Ulrich or be covered here? (the man
page should also be updated to reflect this open flag means append only).
> Fixed
> -----
> Incredibly slow loopback tcp bug (believed fixed about 2.3.48)
Hmm, playing sound via esd and -tcp over loopback is real jerky.
> via rhine oopses under load ?
I've tried to test it but haven't seen any fault with it's current driver.
> Problems with ip autoconfig according to Zaitcev
Didn't he say that it worked now?
> The netdev name changing stuff broke GRE
Incorrect bug. GRE was actually broken by netfilter dropping the packets, not netdev name
changing. This should be fixed in ac11+ iirc. However there is a related GRE bug that is
in the net/ip_gre.c iirc. Rusty and I have patch for it but I haven't checked to see if it
made it into the kernel. Both patches are required for GRE to work. If the second patch is
not applied you'll get protocol 47 unreachables when trying to send packets.
Now for the pcmcia.
With the pcmcia in the kernel, I thought my sockets were #0 on top and #1 on bottom. I
could only use #1 or so I thought. Whenever something was plugged into #0, the kernel would
hang on boot.
Due to my incredible frustration with all tulip cards not working, crashes on eject, crashes
on insert, crashes on boot, basically one workable card in one slot out of about twenty odd
configurations, I had to switch to dhinds pcmcia pkg.
Doing so gave me an clue. Socket #0 is on the bottom and socket #1 is on the top. Both are
perfectly usable. Therefore I suspect an off by one bug in the socket setup. I have not
had time to inspect it further than that.
On the note of the tulips, there are two issues. First, insmod/rmmod causes a reliable OOPS
on the second invocation and both the in core tulip driver and the dhinds tulip driver do
this. Second, when trying to use the pcmcia + tulip in the kernel, no packets are delivered
in or out. Reason? Because interrupts aren't getting delivered. The interrupt count never
increases beyound the first two at initialization and the tulip interrupt handler is never
called.
Using the dhinds package, all my tulip, serial, and wireless cards are useable in either
socket with repeated insert/eject save for the tulips.
There is one problem with the ray_cs driver that is notable, if it is suspended for a long
period of time, it won't wake up.
-d
-- "The difference between 'involvement' and 'commitment' is like an eggs-and-ham breakfast: the chicken was 'involved' - the pig was 'committed'."
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:30 EST