>>>>> "Sergey" == Sergey Vlasov <vsu at altlinux.ru> writes: Sergey> Actually the situation is worse. This driver pokes at SuperIO Sergey> configuration registers, which are shared by all logical Sergey> devices of the SuperIO chip. Exactly (this is something we discussed on a French mailing-list). However, in our case (dedibox.fr hardware), that was not an issue because we only use single-core boards, no other parts of the SuperIO are used at the same time and every existing code sets the logical device number to use each time it needs to access it. Also note that this situation exists with any Winbond SuperIO already in the kernel, this is not specific to this one. A SuperIO subsystem would be the cleanest solution IMO, even if at the beginning it only provides lock/unlock functionality. Then it could be extended as to factor common operation (selecting a logical device and reading/writing registers) and to allow access to the device from userland. On an unrelated subject, the whole watchdog subsystem would benefit from a refactoring; the code looking for a 'V' in a write() operation is present at least 35 times. Sam -- Samuel Tardieu -- sam at rfc1149.net -- http://www.rfc1149.net/
