USB settling time

July 14, 2010 — BarryK

I am posting this link here, as it bothers me and I might want to revert that patch:

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=a4a47bc03fe520e95e0c4212bf97c86545fb14f9

This is for the 2.6.34 kernel. Linus reduced the settling time when detecting USB devices from 5 to 1 second.

This bothers me because there have been some situations in the past where considerable settling time was required. That may have been for USB1 interfaces. Is Linus basically saying that USB1 is no longer being supported?

I don't know, but I thought it wise to flag this patch, so if we encounter any trouble with USB on older hardware, then we will be reminded what the cause might be and the cure.

Linus's patch:

diff --git a/drivers/usb/storage/usb.c b/drivers/usb/storage/usb.c

index e9f9954..bbeeb92 100644 (file)

--- a/drivers/usb/storage/usb.c

+++ b/drivers/usb/storage/usb.c

@@ -78,7 +78,7 @@ MODULE_AUTHOR("Matthew Dharm <mdharm-usb@one-eyed-alien.net>");

 MODULE_DESCRIPTION("USB Mass Storage driver for Linux");

 MODULE_LICENSE("GPL");

 

-static unsigned int delay_use = 5;

+static unsigned int delay_use = 1;

 module_param(delay_use, uint, S_IRUGO | S_IWUSR);

 MODULE_PARM_DESC(delay_use, "seconds to delay before using a new device");

Comments

Fate of USB 1
Username: Sage
It would be extremely helpful if someone respected in the IT community make it abundantly clearly to whoever that dropping support for USB 1 is and will remain utterly unacceptable. Far too much USB 1 kit around in perfect working order, we don't want more landfill. "14 Jul 2010, 15:29"01723"78.149.237.153'1 second is short"JustGreg"I have a USB hard drive attach to my system (with hard disk) for experiments and testing. Lucid takes from 3 to 4 seconds for the usbstorage module to be load. I suspect that 1 second settling time will cause a problem. Sage, the patch is not dropping support for USB 1.0 or 1.1 services. Older devices will be detected if they respond quickly enough."14 Jul 2010, 20:13"01723"63.240.228.37'kernel debugging process"technosaurus"There have been quite a few process speedups (some ~80-90%) that may have made the 5s time unnecessarily long. I like that Linus went straight to 1s so that any problems get identified more quickly in development rather than slowly scaling it back over several releases and delaying the pleasures of a fast boot. ... Does anyone else have a Dectop (usb1 only) that they plan on testing?"14 Jul 2010, 22:18"01723"138.162.8.58'ipods ?"linuxcbon"See http://kerneltrap.org/mailarchive/git-commits-head/2010/2/26/23478/thread Username: 15 Jul 2010, 2:32
"01723"93.25.50.30'usb kernel issues"01micko"I have compiled 2.6.35.13 lts kernel with view of using it in spup. I have added iguleder's name patch so that it appears as 2.6.35. It works great except that the usb-storage patch (I got yours for 2.6.35.8) crashes the kernel even on a boot from CD. I wouldn't have a clue why. It boots without the patch on my lowly eee-701 via USB stick fine.(same with 2.6.37.6 too). It won't boot on a compaq laptop I have from 2008, 1.8gig celeron, 1gig RAM. Since the usb-storage patch fails do you think the patch referred to in the OP would be of any use? Note that all aufs patches apply cleanly and other patches including vortex (though I can't test that it works).

Now compiling 2.6.39.3
Username: BarryK
"I'll see what I can do. Right now I am compiling 2.6.39.3, without the usb-patch and with usb drivers builtin (not as modules) -- this includes usbcore, ehci-hcd, uhci-hcd, ohci-hcd and usb-storage. I have modified the 'init' script to work with these drivers builtin. However, the settling time for usb devices to become available still has to be resolved, but I am hoping that having the drivers builtin might help.

2.6.39.3 kernel
Username: pemasu
"If you get working and usb booting kernel made of the 2.6.39.3, could you provide the patched source. I am keenly interested to test it in the puppy build using your latest woof.

Tags: puppy