Bug 10259
| Summary: | /sys/class/hwmon/hwmon0 is missing a device link | ||
|---|---|---|---|
| Product: | Drivers | Reporter: | Jean-Luc Coulon (jean.luc.coulon) |
| Component: | Hardware Monitoring | Assignee: | Jean Delvare (jdelvare) |
| Status: | CLOSED CODE_FIX | ||
| Severity: | normal | CC: | bunk, rui.zhang |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.25-rc5-git4 | Subsystem: | |
| Regression: | Yes | Bisected commit-id: | |
| Bug Depends on: | |||
| Bug Blocks: | 9832 | ||
| Attachments: | Ignore hwmon devices with no device link | ||
With git4, what are the contents of the "name" attribute? Don't you have an additional "hwmon1" device with git4? I suspect that "hwmon0" is the new ACPI thermal zones virtual device with git4 and you have "hwmon1" which is equivalent to git3's "hwmon0". [jean-luc@tangerine] % cat /sys/class/hwmon/hwmon0/name
thermal_sys_class
Yes, hwmon0 contents has been moved to hwmon1 (and hwmon1 to hwmon2).
Here is the whole tree for git2 ("name" attribut doesnt exist in git3) :
/sys/class/hwmon/
|-- hwmon0
| |-- device -> ../../../devices/pci0000:00/0000:00:18.3
| |-- power
| | `-- wakeup
| |-- subsystem -> ../../hwmon
| `-- uevent
`-- hwmon1
|-- device -> ../../../devices/platform/w83627hf.656
|-- power
| `-- wakeup
|-- subsystem -> ../../hwmon
`-- uevent
and the whole tree with git4:
/sys/class/hwmon/
|-- hwmon0
| |-- name
| |-- power
| | `-- wakeup
| |-- subsystem -> ../../hwmon
| `-- uevent
|-- hwmon1
| |-- device -> ../../../devices/pci0000:00/0000:00:18.3
| |-- power
| | `-- wakeup
| |-- subsystem -> ../../hwmon
| `-- uevent
`-- hwmon2
|-- device -> ../../../devices/platform/w83627hf.656
|-- power
| `-- wakeup
|-- subsystem -> ../../hwmon
`-- uevent
Created attachment 15301 [details] Ignore hwmon devices with no device link This patch to lm-sensors 3.0.x works around the problem by ignoring hwmon devices with no device link. I'll discuss the actual fix with the author of the new generic thermal zone driver. A similar patch has already been applied to the lm-sensors 2.10.x tree: http://www.lm-sensors.org/changeset/5147 I've rebuilt the lm-sensor package after applying the patch. Everything went fine. Thanks Jean-Luc This entry is being used for tracking a regression from 2.6.24. Please don't close it until the problem is fixed in the mainline. Handled-By : Jean Delvare <khali@linux-fr.org> Patch : http://bugzilla.kernel.org/attachment.cgi?id=15301&action=view The problematic patch has been reverted from mainline: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=91f57fa1215952af579f304821a15cc6c656042c We now have some time to discuss how to implement this feature in 2.6.26, and to update libsensors accordingly. |
Latest working kernel version: 2.6.25-rc5-git3 Earliest failing kernel version: 2.6.25-rc5-git4 Distribution: Debian/Sid Hardware Environment: ASUS A8V, Atnlon 64 x2 4200+ Software Environment: Problem Description: "sensors" command returns "sensors_init: Kernel interface error" while the correct modules are loaded Steps to reproduce: boot the system, run snesors or libsensors related applications With git3, I've the following tree: /sys/class/hwmon/hwmon0/ |-- device -> ../../../devices/pci0000:00/0000:00:18.3 |-- power | `-- wakeup |-- subsystem -> ../../hwmon `-- uevent with git4, I get: /sys/class/hwmon/hwmon0/ |-- name |-- power | `-- wakeup |-- subsystem -> ../../hwmon `-- ueven "sensors" complains about the missing device link: <snip> open("/sys/class/hwmon", O_RDONLY|O_NONBLOCK|O_DIRECTORY|0x80000) = 3 fstat(3, {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0 getdents(3, /* 5 entries */, 4096) = 144 readlink("/sys/class/hwmon/hwmon0/device", 0x7fffe7b98f60, 254) = -1 ENOENT (No such file or directory) close(3) = 0 write(2, "sensors_init: Kernel interface e"..., 37sensors_init: Kernel interface error ) = 37 <snip/>