cvs commit: src/sys/dev/usb usb.c

[Ian Dowse]

In message <200503252051.aa57514 at salmon.maths.tcd.ie>, Ian Dowse writes:
>In message <424473BB.4010607 at portaone.com>, Maxim Sobolev writes:
>>This is still suboptimal, since even after this change when ehci is 
>>enabled some (if not most) USB devices are detected after system goes 
>>into milti-user mode. Obviously, this makes it impossible to mount root 
>>fs off the USB HDD/FDD/CD. The simplest way to solve this problem is to 
>>add some sufficiently long and configurable delay before mounting root 
>>fs (similar to one used by the SCSI) when ehci is enabled to let those 
>>devices to be detected and attached.
>
>Could you send me an example dmesg that shows this? As far as I
>know, doing the explore when cold == 1 should attach all USB devices
>before even interrupts are enabled.
>
>This should be the case with PCI USB host controllers that are
>compiled into the kernel or pre-loaded by the loader - if you load
>the module later and I think also with cardbus USB controllers,
>then the USB host controller won't get attached with cold == 1 so
>none of the early exploration code runs.

Booting from a USB disk and using it as the root filesystem works
on my laptop anyway:

	sc0: VGA <16 virtual consoles, flags=0x300>
	vga0: <Generic ISA VGA> at port 0x3c0-0x3df iomem 0xa0000-0xbffff on isa0
	umass0: USB Flash Disk, rev 2.00/2.00, addr 2
	Timecounter "TSC" frequency 1000038929 Hz quality 800
	Timecounters tick every 1.000 msec
	ad0: 38154MB <IC25N040ATMR04-0/MO2OAD4A> [77520/16/63] at ata0-master UDMA100
	da0 at umass-sim0 bus 0 target 0 lun 0
	da0: <JetFlash TS256MJF2B 2.00> Removable Direct Access SCSI-2 device 
	da0: 40.000MB/s transfers
	da0: 250MB (512000 512 byte sectors: 64H 32S/T 250C)
	Trying to mount root from ufs:/dev/da0s1a
	Pre-seeding PRNG:

Ian