Hello, in case you didn't notice, here's the device-mapper vs. sysfs topic coming up again: -------- Weitergeleitete Nachricht -------- Von: W. Michael Petullo <mike@xxxxxxxx> An: linux-kernel@xxxxxxxxxxxxxxx Betreff: Dm-crypt, device-mapper and sysfs Datum: Mon, 1 Nov 2004 13:52:00 -0600 I am interested in allowing the Hardware Abstraction Layer (HAL) to interface with dm-crypt devices. I would like to see hald have the ability to automatically mount removable encrypted drives when they are plugged in. I have a working prototype of this idea except for one issue. Currently, dm-crypt registers its managed devices in the kernel's sysfs. This is good. However, unlike most of the devices in /sys, dm-crypt devices don't have a link named device. For most devices, this link specifies the bus the device is connected to. On my iBook, for example, /sys/block/hda/device points to /sys/devices/pci0002:02/0002:02:0d.0/ide0/0.0. This is my IDE bus. Hald expects to find a link like this when it is building its device tree. I've discussed this issue on the dm-crypt mailing list and Christophe thinks this is a more general device-mapper issue. So I don't put words in his mouth, here is what he said: > dm-crypt is a device-mapper target, it doesn't do any sysfs interaction. > There were some plans, even patches, to export more device-mapper > functionality into sysfs. Besides, a mapped device can consist of > several other devices and different targets. cryptsetup has the ability > to query an existing device, it will tell you if it is a "valid > cryptsetup'ed device". Also there a are plans to put some metadata into > the first blocks of the device so that it can be automatically > recognized as encrypted device. So my question is this: Should the device-mapper driver be modified to provide a device link? If so, what should it point to?
Attachment:
signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil