- Class io_res
- krishna: OSKit AMM alternative?
- Member simple_strtol
- krishna: base is `unsigned int' NOT `int', but it's always `0' in the code (as far as I know)
- Member io_loop
- krishna: Hmm, is it clever to reference io_self in request?
- Member io_loop
- krishna: Do we need receive timeouts?
- Member l4_io_unregister_client_component
- krishna: At this point we have to keep track of up-to-date resource allocation info -> all allocated resources should be freed on unregistering. (just I don't mind about this at this state, later driver replacing will be in scope of this)
- Member __request_resource
- krishna: it seems as Linux only sorts resources here; exception: PCI conf #1 I/O ports
- Member __request_resource
- krishna: now: top-level announcement and BUSY requests
- Member __request_region
- krishna: As Linux grabs a write_lock(resource_lock), we assume single-threading.
- Member request_resource
- krishna: As Linux grabs a write_lock(resource_lock), we assume single threading.
- Member request_resource
- krishna: Try to align MMIO regions before doing real allocation (see also allocate_resource()).
- Member allocate_resource
- krishna: As Linux grabs a write_lock(resource_lock), we assume single-threading.
- Member kmalloc
- krishna: Okay, PCIlib never kmallocs with gfp != GFP_KERNEL, so malloc seems sufficient.
- Member FASTCALL
- krishna: It's called by
pci_alloc_consistent()
and pcibios_get_irq_routing_table()
- so do we _really_ need it?
- Member FASTCALL
- krishna: It's called by
pci_free_consistent()
and pcibios_get_irq_routing_table()
- so do we _really_ need it?
- Member callback_announce_mem_region
- krishna: Hopefully we'll get no doubled announcements - I don't check.
- Member io_res_init
- krishna: This list is not complete ...
l4io, written by
Christian Helmuth
© 2003 Technische Universitaet Dresden