Re: Regression in v4.2-rc1: vmalloc_to_page with ioremap

[Toshi Kani <toshi.kani@hp.com>]

On Mon, 2015-07-20 at 09:54 -0600, Toshi Kani wrote:
> On Sat, 2015-07-18 at 18:04 +0000, Dixit, Ashutosh wrote:
> > vmalloc_to_page with ioremap'd memory used to work correctly till 
> > v4.1. In
> > v4.2-rc1 when ioremap is done using huge pages vmalloc_to_page on 
> > ioremap'd
> > memory crashes. Are there plans to fix this?
> > 
> > An example of the use of vmalloc_to_page with ioremap is in the 
> > Intel 
> > MIC SCIF
> > driver (drivers/misc/mic/scif/scif_nodeqp.c) which dma map's the 
> > ioremap'd range
> > of one device to a second device.
> 
> Yes, vmalloc_to_page() assumes 4KB mappings.  But ioremap with huge
> pages was enabled in v4.1, while v4.2-rc1 has update for the check 
> with
> MTRRs.  Can you send me outputs of the following files?  If the 
> driver
> fails to load in v4.2-rc1, you can obtain the info in v4.1.
> 
>   /proc/mtrr
>   /proc/iomem
>   /proc/vmallocinfo
>   /sys/kernel/debug/kernel_page_tables (need CONFIG_X86_PTDUMP set)
> 
> Also, does the driver map a regular memory range with ioremap?  If 
> not, how does 'struct page' get allocated for the range (since
> vmalloc_to_page returns a page pointer)?

Can you also try the 'nohugeiomap' kernel option to the 4.2-rc1 kernel
to see if the problem goes away?

Thanks,
-Toshi