Greets, fellow Slackers. Got a weird issue. Slack 10, custom 2.6.12.2 kernel, Nforce2 chipset, ehci_hcd & ohci_hcd modules loaded. I plug in the Sandisk MP3 player, and here is the dmesg

usb 3-2: new high speed USB device using ehci_hcd and address 5
scsi9 : SCSI emulation for USB Mass Storage devices
usb-storage: device found at 5
usb-storage: waiting for device to settle before scanning
Vendor: SanDisk Model: SDMX1 MP3 Player Rev: 1.00
Type: Direct-Access ANSI SCSI revision: 00
SCSI device sda: 497664 512-byte hdwr sectors (255 MB)
sda: Write Protect is off
sda: Mode Sense: 23 00 00 00
sda: assuming drive cache: write through
SCSI device sda: 497664 512-byte hdwr sectors (255 MB)
sda: Write Protect is off
sda: Mode Sense: 23 00 00 00
sda: assuming drive cache: write through
sda: sda1
Attached scsi removable disk sda at scsi9, channel 0, id 0, lun 0
usb-storage: device scan complete

Happy camper right? Then I mount it as /mnt/usb-mp3 with options users,rw,sync. And it mounts just fine.

Here's the weird part. When I transfer files from the MP3 player to the harddrive, I get speeds of over 7MB/s. Nice. But when I transfer files from HD to MP3 player, my speeds drop 250KB/s. What gives?! Nothing in dmesg. I've tried reloading all the modules and remounting. I removed the sync mount option, but that has it's own quirks and I don't want to do down that path.

Any suggestions?

--Akshun J

Hmmm, after reviewing /var/log/messages, I see that I've got a FATAL: Module sd_mod not found.

Who wants to bet that's my problem... Grrrrr...


--Akshun J

Well, my kernel config says

CONFIG_BLK_DEV_SD=y

This appears to be compiled in, so no wonder the module loading failed. It should be working, though. I'm really at a loss for figuring this out now...

--Akshun J

7MB/s is faster than I've ever seen my 2.0 ports go (averages 5 with my external USB HDD), but then again, I get 5 both ways... I'm not familiar with the device in question, and even though I doubt it's the case, is it possible that the hardware on its end just isn't capable of faster writes than that? Just a thought...
Also, I do recall seeing in the kernel config help a somewhat cryptic blurb about a particular USB block device (if memory serves) driver/option that said something like 'be aware enabling this cripples the USB driver' -- what I know I've picked up by trial and error alone, so my info's shaky, but perhaps you've compiled in that option and the sluggish output is the result that they were hinting at (which if I remember how I configured mine, I included as a module and have no probs with)?
I've had odd instances of my system trying to load modules that it shouldn't be trying to find in the first place, as well; wish I could be more helpful on that, but is there any chance you missed a soft link when compiling your 2.6 kernel and Slack is still working as though you're using it's 2.4? (config files in /etc/rc.d/ dir, do a fresh "make && make modules_install" in the 2.6 kernel's src dir, update the soft links in /lib/modules and /usr/src to point to the current version incase it isn't just for user convenience, etc.)
Might try using the same config and just recompiling and copying the system.map and bzImage files to your /boot dir and checking lilo.conf as a last resort, I suppose.
Good luck figuring it out

Maybe it's the sync option that is slowing you down. Do you have the same issue if you drop the sync option?

I tried dropping the sync option, and it helps, but creates other issues. I copy something, it completes, and then when I unmount the device it takes a couple of minutes while it syncs the data. Not a viable option. Thanks, though.

--Akshun J

I think that's just the nature of the beast. The sync option makes transfers happen one at a time. Dropping the sync option just does everything at once, not waiting for the last action to complete. You may have to wait a few moments after unmounting, but I have found that the wait is NOWHERE NEAR as long as the waitng around for file transfers with the sync option.

It's the "Low Performance USB Block driver", I also encountered USB problem because of this module.

So akshunj you can try to remove this module and see if it resovle your USB problem.

2 months later...

It was not the low performance USB driver. I NEVER compile that thing in. It's useless...

I have since upgraded to 2.6.13.2, but I haven't tested this function yet. I'll update the thread when I do...

--Akshun J