vm_soft_fault when booting with pxe and a haiku.image through remote_disk_server,

[oco]

When booting a machine with pxe and a haiku.image through the remote_disk_server, there is a vm_soft_fault. The system seems to reload the network stack, already used to access the remote disk.

Here is the error :

vm_soft_fault: va 0x8164a18a not covered by area in address space vm_page_fault: vm_soft_fault returned error 'Bad address' on fault at 0x8164a18a , ip 0x8164a18a, write 0, user 0, thread 0x2d PANIC: vm_page_fault: unhandled page fault in kernel space at 0x8164a18a, ip 0x8 164a18a.

The serial output is attached (with stack trace).

The computer is an old celeron 300 with a rtl8139 network card. I use a gPXE floppy disk to boot the system.

Workaround :

• mount the haiku.image
• rename /<mount point>/beos/system/add-on/kernel/drivers/dev/net to /<mount point>/beos/system/add-on/kernel/drivers/dev/net.arch.
• unmount

[korli]

Did it work before ? There is also a similar problem in #1909, as if the devfs mixed up vnodes, cookies or hooks when republishing a driver.

[oco]

I don't know ! This is the first time i boot an image through remote_disk_server...

[oco]

Fixed !

Netboot now work fine on my ten years old Celeron 300 Mhz with 192 Mo of RAM using a gPXE floppy ! I am currently posting from this good old friend !

Tested with :

• pre-alpha image from hrev29155,
• haiku-netboot.tgz and pxehaiku-loader from hrev28998 (available here ​http://olivier.coursiere.free.fr/download/r28998/).

[korli]

Thanks for the feedback!