[Acx100-devel] [PATCH 00/14]

[Jim C. <jim...@gm...>]

Hi Oliver, Larry,

Another batch:

1- revisit debugfs.c typecasts again.  I used size_t, since its an
   integer type (like fidx), and changes size with platform (like
   uintptr_t)

2- corrects broken patch from previous batch
3- fixes incomplete patch which Oliver had to fix up for 2.6.32-3
   hopefully this finishes it.

4- candy
5- DECL_OR_STUB reduces double prototypes

10- I think acxpci-upload-fw has some cruft that could be removed,
    after you (Oliver) fixed the separate firmware loading in
    _acx_upload_fw, needed by mem

0001-debugfs.c-fix-types-warnings.patch
0002-common.c-set-debug-flags-by-symbolic-names.patch
0003-abstract-ieee80211_VIF-changes-in-acx_op_-add-remove.patch
0004-common.c-replace-if-then-else-with-min-in-switch-cas.patch
0005-add-DECL_OR_STUB-macro-to-acx.h-use-it-in-merge.h.patch
0006-pci.h-drop-acxpci_upload_fw-prototypes.patch
0007-drop-STATick-no-longer-needed.patch
0008-cleanup-ifdefs-in-common.c-init-exit-funcs.patch
0009-merge.c-make-acx_process_rxdesc-static.patch
0010-cleanup-upload_fw.patch
0011-add-MACSTR_SIZE.patch
0012-hide-acx_op_tx-iface-change-in-acx_compat.h.patch
0013-cleanup-debugfs.c-fn-protos.patch
0014-scripts-wifi-probe.rc-loads-wifi-probe.prefs.patch

acx-mac80211$ git diff --stat master
 acx.h                 |  61 ++++++++++++++++++++++++++-
 acx_compat.h          |  32 +++++++++++---
 acx_func.h            |   1 +
 acx_struct_dev.h      |   2 +-
 common.c              | 113 ++++++++++++++++----------------------------------
 debugfs.c             |  31 ++++++--------
 mem.c                 |  27 ++++++------
 mem.h                 |   3 --
 merge.c               |  10 ++---
 merge.h               |  74 ++++++++++++++++-----------------
 pci.c                 |  88 +--------------------------------------
 pci.h                 |  16 ++++---
 scripts/wifi-probe.rc |  19 ++++++++-
 usb.c                 |   4 +-



Other Items:

Im certain that debugfs files livecycle is not quite correct.  In
addition to expected /dbg/acx_mac80211/phyX, Im seeing /dbg/phyX show
up also (on ifup I think), and theyre not going away on ifdown.

I cleaned up PROC_FS vs DEBUG_FS a bit, but theres more to be done.
EXTRA_CFLAGS:= -g -DACX_WANT_PROC_FILES_ANYWAY \ breaks currently.
I think this one simplifies once lifecycle issues are fixed.

your __acquire/__release patch seems to fix sparse complaints.
Of course, its now thowing new ones.

  CHECK   /home/jimc/projects/lx/acx-mac80211/merge.c
merge.c:463:48: warning: cast removes address space of expression
merge.c:563:20: warning: cast removes address space of expression
merge.c:1618:32: warning: incorrect type in assignment (different address spaces)
merge.c:1618:32:    expected unsigned char [noderef] [usertype] <asn:2>*[usertype] cmd_area
merge.c:1618:32:    got unsigned char [usertype] *<noident>
merge.c:1619:33: warning: incorrect type in assignment (different address spaces)
merge.c:1619:33:    expected unsigned char [noderef] [usertype] <asn:2>*[usertype] info_area
merge.c:1619:33:    got unsigned char [usertype] *<noident>
interrupt-masks.h:10:5: warning: symbol 'interrupt_masks' was not declared. Should it be static?
interrupt-masks.h:99:12: warning: symbol 'devtype_names' was not declared. Should it be static?
interrupt-masks.h:100:12: warning: symbol 'chiptype_names' was not declared. Should it be static?
interrupt-masks.h:109:29: warning: symbol 'interrupt_descs' was not declared. Should it be static?
interrupt-masks.h:165:6: warning: symbol 'interrupt_show_flags' was not declared. Should it be static?
include/linux/spinlock.h:340:9: warning: context imbalance in 'acx_reset_dev' - unexpected unlock
interrupt-masks.h:33:25: warning: cast truncates bits from constant value (ffff10fc becomes 10fc)
  CHECK   /home/jimc/projects/lx/acx-mac80211/debugfs.c


Im mildly concerned about playing with lock *hooks*

linux-2.6$ more Documentation/lockstat.txt 

LOCK STATISTICS

<ascii art>

lock, unlock	- the regular lock functions
__*		- the hooks
<> 		- states

With these hooks we provide the following statistics:

 con-bounces       - number of lock contention that involved x-cpu data
 contentions       - number of lock acquisitions that had to wait
 wait time min     - shortest (non-0) time we ever had to wait for a lock
           max     - longest time we ever had to wait for a lock
           total   - total time we spend waiting on this lock
 acq-bounces       - number of lock acquisitions that involved x-cpu data
 acquisitions      - number of times we took the lock
 hold time min     - shortest (non-0) time we ever held the lock
           max     - longest time we ever held the lock
           total   - total time this lock was held


I just rebuilt my kernels with CONFIG_LOCK_STAT, to see if theres
anything surprising in there, will try driver before/after the patch,
idle/under-load, and let you know what I see.