On Wed, 30 Aug 2006 11:32:47 -0600
"Erik Habbinga" <[email protected]> wrote:
> The mpt_downloadboot function in drivers/message/fusion/mptbase.c doesn't work correctly on big endian systems (powerpc in my case).
> I've added appropriate le32_to_cpu calls to correctly translate little endian firmware file data into cpu endian format before
> getting written to little endian PCI registers. This patch has been tested successfully on a powerpc target and an Intel target.
>
> Signed-off-by: Erik Habbinga <[email protected]>
>
> --- a/drivers/message/fusion/mptbase.c.orig 2006-08-23 15:16:33.000000000 -0600
> +++ b/drivers/message/fusion/mptbase.c 2006-08-30 10:28:39.000000000 -0600
> @@ -2915,7 +2915,7 @@
> u32 ioc_state=0;
>
> ddlprintk((MYIOC_s_INFO_FMT "downloadboot: fw size 0x%x (%d), FW Ptr %p\n",
> - ioc->name, pFwHeader->ImageSize, pFwHeader->ImageSize, pFwHeader));
> + ioc->name, le32_to_cpu(pFwHeader->ImageSize), le32_to_cpu(pFwHeader->ImageSize), pFwHeader));
>
> CHIPREG_WRITE32(&ioc->chip->WriteSequence, 0xFF);
> CHIPREG_WRITE32(&ioc->chip->WriteSequence, MPI_WRSEQ_1ST_KEY_VALUE);
> @@ -2968,7 +2968,7 @@
> /* Set the DiagRwEn and Disable ARM bits */
> CHIPREG_WRITE32(&ioc->chip->Diagnostic, (MPI_DIAG_RW_ENABLE | MPI_DIAG_DISABLE_ARM));
>
> - fwSize = (pFwHeader->ImageSize + 3)/4;
> + fwSize = (le32_to_cpu(pFwHeader->ImageSize) + 3)/4;
> ptrFw = (u32 *) pFwHeader;
>
> /* Write the LoadStartAddress to the DiagRw Address Register
> @@ -2977,23 +2977,23 @@
> if (ioc->errata_flag_1064)
> pci_enable_io_access(ioc->pcidev);
>
> - CHIPREG_PIO_WRITE32(&ioc->pio_chip->DiagRwAddress, pFwHeader->LoadStartAddress);
> + CHIPREG_PIO_WRITE32(&ioc->pio_chip->DiagRwAddress, le32_to_cpu(pFwHeader->LoadStartAddress));
> ddlprintk((MYIOC_s_INFO_FMT "LoadStart addr written 0x%x \n",
> - ioc->name, pFwHeader->LoadStartAddress));
> + ioc->name, le32_to_cpu(pFwHeader->LoadStartAddress)));
>
> ddlprintk((MYIOC_s_INFO_FMT "Write FW Image: 0x%x bytes @ %p\n",
> ioc->name, fwSize*4, ptrFw));
> while (fwSize--) {
> - CHIPREG_PIO_WRITE32(&ioc->pio_chip->DiagRwData, *ptrFw++);
> + CHIPREG_PIO_WRITE32(&ioc->pio_chip->DiagRwData, le32_to_cpu(*ptrFw++));
hm, I'd be a bit more comfortable if this didn't require that le32_to_cpu()
was sanely implemented. include/linux/byteorder/swab.h has
# define __swab32(x) \
(__builtin_constant_p((__u32)(x)) ? \
___swab32((x)) : \
__fswab32((x)))
so if `x' is foo++, what value does `foo' end up with? I guess it would be
pretty dumb if __builtin_constant_p(foo++) were to increment foo, but is
that guaranteed?
Anyway. This is a moderately important fix, isn't it? Eric (Moore), do
you want to proceed with this patch?
Thanks.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]
