Differences

This shows you the differences between two versions of the page.

--- soc:2010:peper:journal:week10 [2010/07/26 15:44]
peper
+++ soc:2010:peper:journal:week10 [2010/07/31 15:13] (current)
peper
@@ Line 3: / Line 3: @@
 ==== drivers in userspace ====
+=== The big issue ===
 There is one thing that keeps bugging me in the [[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/drivers|drivers branch]] - UIO-DMA and malloc.
 UIO-DMA does the DMA mappings on a per device basis (as does kernel). I am
@@ Line 16: / Line 17: @@
 around that I have added an option to use the last device setup with
 UIO-DMA for the mappings and introduced a UIO-DMA malloc_backend - [[http://git.etherboot.org/?p=people/peper/gpxe.git;a=commitdiff;h=14009a8a96431cecd089c70d6ac44bfe75244497|here]] and [[ http://git.etherboot.org/?p=people/peper/gpxe.git;a=commitdiff;h=941ffdafdac66bcbca04fbac1668a4b944c7c18a|here]].
-The reason I use tables (and not single api) for the backends
+The reason I use the tables API (and not single API) for the backends
 is that it needs to be able to fallback so that using the tap driver
-is still possible w/o the modules.
+is still possible without the UIO-DMA modules being loaded and initialized).
 I don't really like the solution and I am starting to think that doing
@@ Line 28: / Line 29: @@
 code size.
-Or I could make the malloc
+Or I could make the malloc backend a single API (no need for fallback any more) and implement it in userspace like that:
-backend a single api (no need for fallback any more) and for userspace
+  * in a __init_fn provide the malloc pool by mmap()ing a chunk of memory (like linux_umalloc does) - let's call it A
-on init mmap a chunk of memory (like linux_umalloc does) and then
+  * if an lpci device is going to be used:
-during lpci init allocate the same amount with UIO-DMA, copy the old
+     * allocate the same amount of memory as for A from UIO-DMA - let's call the new chunk B
-memory over, unmap the old chunk and remap the newly allocated chunk
+     * copy A over to B
-in that place.
+     * munmap() A (this isn't really needed as mremap() can take care of that)
-I like the last option most cause it uses the current malloc()
+     * mremap() B to the now free location of A
-implementation, works for tap and doesn't feel so hackish. I am going to implement that soonish unless other ideas arise :)
+I like the last option most cause it uses the current malloc() implementation, works for tap and doesn't feel so hackish (just sophisticated ;). I am going to implement that soonish unless other ideas arise.
+== Update ==
+"Sophisticated" didn't really work for Josh, but I have come up with something different. I have introduced separate memery pools for normal and DMA memory allocation with an API for switching the latter.
+See [[http://git.etherboot.org/?p=people/peper/gpxe.git&a=search&h=refs/heads/drivers&st=commit&s=[malloc]+Introduce+memory+pools+and+hide+internal+API|[malloc] Introduce memory pools and hide internal API]]
+=== Smaller issues ===
+  * currently ''gpxe.linux'' binary doesn't contain the ''tap'' driver - easily fixable
+  * out/in* segfault in userspace if ''iopl()'' wasn't called - will probably have to check whether the ioports were initialized on each call
+== Update ==
+These have been fixed. Moreover, slighty related to the first one, I have come up with a [[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/buildall|buildall branch]],
+ which allows building alldrivers builds on all supported arch/platform combinations and also adds a everything target that takes advantage of that and builds everything that I could think of. Should come in handy for testing patches.
+The changes are mostly trivial, but if you want to refresh your make-foo have a look at [[http://git.etherboot.org/?p=people/peper/gpxe.git&a=search&h=refs/heads/buildall&st=commit&s=[build]+Properly+handle+multiple+goals+per+BIN+directory|[build] Properly handle multiple goals per BIN directory]].
+=== Update ===
+I have updated pretty much every commit in the [[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/drivers|drivers branch]], adding comments and doing cleanup.