summaryrefslogtreecommitdiff
path: root/sys/src/9/omap/notes
diff options
context:
space:
mode:
authorTaru Karttunen <taruti@taruti.net>2011-03-30 15:46:40 +0300
committerTaru Karttunen <taruti@taruti.net>2011-03-30 15:46:40 +0300
commite5888a1ffdae813d7575f5fb02275c6bb07e5199 (patch)
treed8d51eac403f07814b9e936eed0c9a79195e2450 /sys/src/9/omap/notes
Import sources from 2011-03-30 iso image
Diffstat (limited to 'sys/src/9/omap/notes')
-rwxr-xr-xsys/src/9/omap/notes/movm.w22
1 files changed, 22 insertions, 0 deletions
diff --git a/sys/src/9/omap/notes/movm.w b/sys/src/9/omap/notes/movm.w
new file mode 100755
index 000000000..a2b22f1e1
--- /dev/null
+++ b/sys/src/9/omap/notes/movm.w
@@ -0,0 +1,22 @@
+gorka writes:
+---
+I have userspace on the gumstix [xscale, not omap]. The problem that
+got me in trouble was that in lexception.s (or l.s),
+
+ MOVM.DB.W [R0-R14], (R13)
+
+works differently for this architecture (and probably for others, as
+it is unclear how it should behave by reading the arm specs). This
+happens only for kernel faults as the others (syscall, user faults)
+use MOVM.DB.W.S which uses the banked user registers.
+
+The problem is that in this arch the value of R13 saved is the value
+after R13 itself has been modified, whereas in the others (bitsy,
+pico...), it was the value before. Adding 4*15 to the stack before
+the RFE solves the problem.
+---
+
+In fact, the 2005 ARM arch. ref. man. (ARM DDI 0100I) says, under STM (1),
+that if Rn appears in the set of registers (and isn't the first one)
+and .W is specified, the stored value of Rn is unpredictable.
+The arm v7-ar arch. ref. man. says such usage is obsolete.