Red Hat Bugzilla – Bug 135886
Reading data off block device from USB memory stick fails
Last modified: 2013-03-05 22:41:49 EST
Description of problem:
Reading data off a block device from a USB memory stick fails. This
means that the hal daemon cannot probe for file system type, volume
labels, UUID and so forth. Reading data off IDE devices appear to work.
Version-Release number of selected component (if applicable):
This worked with earlier kernels, I've just tested with .610 and .584.
Steps to Reproduce:
1. Start up hal daemon
2. Plugin a USB stick
3. Verify with hal-device-manager that volume.fstype, volume.label
Will attach strace and debug output from hald from runs on the .624
and the .610 kernel
Created attachment 105284 [details]
Output from hald on .624 when plugging in USB memory stick
Created attachment 105285 [details]
Selected strace output from hald on .624 when plugging in USB memory stick
Created attachment 105286 [details]
Output from hald on .584 when plugging in USB memory stick
Created attachment 105287 [details]
Selected strace output from hald on .584 when plugging in USB memory stick
The difference seems to be that read(2) on the fd from /dev/sda1 fails
with EAGAIN (aka EWOULDBLOCK). Is this a bug in the application?
I'm here in the Westford office if you need more traces, info etc.
OK, as discussed on IRC the kernel update just exposes already flawed
apps wrt. using O_NONBLOCK. I've got a patch that I will apply to the
hal package to fix this.
Reassigning this bug to hal.
This is fixed in hal-0.4.0-3