Hi, What's the most advisable means of storing struct ata_host objects for later reference? Can they be constructed relatively easily from other, more basic, kernel structures which should exist after the ahci.ko driver has loaded? Briefly, what's prompting this is that I'm writing a character driver to act as a pass through from user space to ahci.ko (actually, a modified version of ahci.ko). One of the mods I've made is a linked list of struct ata_host objects which is used by my character driver to know how many device nodes to instantiate (1 for each port on however many HBAs the LLD is driving). At this location, http://lxr.linux.no/linux +*/drivers/ata/ahci.c#L3230 , I placed a call to a function I added which simply allocates, via kalloc(), a new ata_host structure. I then simply memcpy() the contents of what the driver has dutifully instantiated. I figured that copying the structure at this point was safe because it was fully built. The reason for all of this is so that the character driver can keep track of which ata host owns the port, etc. One of the things my character driver allows is access to HBA and port registers. In my very early drafts, I'm calling pci_get_device(), but this just isn't workable because I don't want the upper layer driver tied to specific hardware. I use the ata_host, and its ata_port, objects to get access to the iomapped memory. Essentially, the idea was that the character driver would pass the ata_host structure to the LLD driver for HBA registers, and an ata_port structure for the port registers. Apparently, however, there's more going on and simply copying (a shallow copy at that) these structures doesn't provide enough security. I caused a kernel oops. Is there a better means of going about this? Andy ��.n��������+%������w��{.n�����{��'^�)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
