From: Jim C. <jim...@gm...> - 2012-04-30 00:42:55
|
On Sun, Apr 29, 2012 at 12:03 PM, Oliver Winker <oli...@ol...> wrote: > Hi Jim, > > The latest series of 6 is applied and pushed ;)! Thanks a lot again! ah, very good, I have another batch to finish and send :-D > it was interesting, because the series revealed and reminded me > of a number of things ;): > > I was puzzled by patch#5, which strangely didn't want to apply to > acx_struct_dev.h. Finally it turned out, that acx_struct_dev.h was > using CRLFs ... (took me some time to find out: vim was suddenly showing > ^M and before he wasn't (he showed them only with mixed CRLF/LF, after I > forced the patch into ... anyway ;). I guess theyre in the code from long ago, Im pretty sure I didnt add them, unless emacs has some crypto-compatibility cruft that did it. > So I fixed that before and then everything applied nicely. > Then there was still some UTF-8 chars in mem.h, which I also fixed at > that occasion. Now a "file *.h *.c" is all "(C source, )?ASCII text". Yes, git send-email caught a bunch of them in a previous set that I ^C'd and corrected, I guess some slipped past me this last batch. The irony is that theyre coming from gcc errors. > Patch#4 was also interesting, because I recognized, that acxpci_reset_mac() > is actually not used with CONFIG_VLYNQ. It works anyhow it seems, but I need to > check that one time. I also fixed there the vlynq build - you see. > > Patch#6: This was certainly an interesting exercise ;)! ... and > it did remind me, that the io/membase part of struct acx_device is > really still messy ... it'll somehow work (at least seems to me), as long > as you don't try to use a mem and pci device on 64bit. > > This needs really onetime to be fixed correctly and with the foreseen types > of the io/dma layer (void __iomem) of the kernel. For mem this means, that > the IO-macros would need to adapted accordingly. > > The OOPs you saw was maybe related to the build not being entirely refreshed. > I checked onetime with pahole (see below), and although the overall size didn't > change, the layout did still change. ooh - I knew of that tool once upon a time, but forgot completely about it. > Apart from that, the type mess also actually also made a type-changes on > membase(2) between pci.o and merge.o. I'm not sure if that can become really > problem, but doesn't look consistent at least. > > I added this that to the TODO ;)! The statement above is stronger and more detailed than the TODO addition. I'll keep it in mind. next batch merges acx_create(tx|rx)_desc_queues(), and extends ACX_GIT_VERSION, ie: jimc@chumly:~/projects/lx/acx-mac80211$ modinfo ./acx-mac80211.ko filename: ./acx-mac80211.ko version: v0.6.0-gv20120426-10-ge3c83b8 Plus some stuff Im still hacking at. If you want the above tonite, let me know. > > Cheers and Thanks again! Oliver ;) > thanks, and youre welcome Jim > pci.o, all configs, before commit: > --- >> struct device * bus_dev; /* 1728 4 */ >> struct pci_dev * pdev; /* 1732 4 */ >> long unsigned int membase; /* 1736 4 */ >> long unsigned int membase2; /* 1740 4 */ >> void * iobase; /* 1744 4 */ >> void * iobase2; /* 1748 4 */ > --- > pci.o, all configs, after commit: > --- >> struct device * bus_dev; /* 1728 4 */ >> struct pci_dev * pdev; /* 1732 4 */ >> long unsigned int membase; /* 1736 4 */ >> void * iobase; /* 1740 4 */ >> long unsigned int membase2; /* 1744 4 */ >> void * iobase2; /* 1748 4 */ > --- > > mem.o, all configs, before commit: > --- >> struct device * bus_dev; /* 1728 4 */ >> struct platform_device * pdev; /* 1732 4 */ >> volatile u32 * membase; /* 1736 4 */ >> long unsigned int membase2; /* 1740 4 */ >> volatile u32 * iobase; /* 1744 4 */ >> void * iobase2; /* 1748 4 */ > --- > mem.o, all configs, after commit: > --- >> struct device * bus_dev; /* 1728 4 */ >> struct platform_device * pdev; /* 1732 4 */ >> volatile u32 * membase; /* 1736 4 */ >> volatile u32 * iobase; /* 1740 4 */ >> long unsigned int membase2; /* 1744 4 */ >> void * iobase2; /* 1748 4 */ > --- > > > merge.o, all configs, before commit: > --- >> struct device * bus_dev; /* 1728 4 */ >> struct platform_device * pdev; /* 1732 4 */ >> volatile u32 * membase; /* 1736 4 */ >> long unsigned int membase2; /* 1740 4 */ >> volatile u32 * iobase; /* 1744 4 */ >> void * iobase2; /* 1748 4 */ > --- > merge.o, all config, after commit: > --- >> struct device * bus_dev; /* 1728 4 */ >> struct platform_device * pdev; /* 1732 4 */ >> volatile u32 * membase; /* 1736 4 */ >> volatile u32 * iobase; /* 1740 4 */ >> long unsigned int membase2; /* 1744 4 */ >> void * iobase2; /* 1748 4 */ > --- > > > > On 04/27/2012 06:34 AM, Jim Cromie wrote: >> Signed-off-by: Jim Cromie <jim...@gm...> >> --- >> merge.c | 2 ++ >> 1 files changed, 2 insertions(+), 0 deletions(-) >> >> diff --git a/merge.c b/merge.c >> index 813cdf0..6fe00fb 100644 >> --- a/merge.c >> +++ b/merge.c >> @@ -2453,6 +2453,7 @@ void acxmem_dealloc_tx(acx_device_t *adev, tx_t *tx_opaque) { >> /* >> * Return an acx pointer to the next transmit data block. >> */ >> +#if 0 // using copy in mem.c >> static u32 acxmem_allocate_acx_txbuf_space(acx_device_t *adev, int count) { >> u32 block, next, last_block; >> int blocks_needed; >> @@ -2512,6 +2513,7 @@ static u32 acxmem_allocate_acx_txbuf_space(acx_device_t *adev, int count) { >> >> return block; >> } >> +#endif >> >> /* >> * Return buffer space back to the pool by following the next pointers until we find > |