Re: [PATCH] arm: Handle device tree memory regions larger than 4GB
On 6 July 2012 20:26, Dave Martin wrote: > On Fri, Jul 06, 2012 at 07:07:35PM +0100, Peter Maydell wrote: >> On 6 July 2012 00:27, Rob Herring wrote: >> > I would just change arm_add_memory to use phys_addr_t for the size >> > param. This ultimately calls memblock functions which use phys_addr_t >> > for sizes. >> >> So I have a patch that does this which basically works. However >> there is a bit I'm not sure about. arm_add_memory() does this: >>bank->size = size & PAGE_MASK; >> >> in an attempt to clear the bottom bits of the size. However, >> since PAGE_MASK is defined as: >> #define PAGE_SIZE (_AC(1,UL) << PAGE_SHIFT) >> #define PAGE_MASK (~(PAGE_SIZE-1)) >> >> PAGE_MASK is a 32 bit unsigned constant and so this & will >> clear the top 32 bits of bank->size. >> >> I'm really not sure what the best way to fix this is; suggestions? > > Maybe something like > > ~(phys_addr_t)~PAGE_MASK > > or > > ~(phys_addr_t)(PAGE_SIZE - 1) > > would work. Mmm. It feels a bit unsatisfactory that an "obviously correct" line of code like "size &= ~PAGE_MASK" could actually be subtly wrong (seems like the kind of thing that could easily slip through code review), so I was wondering if maybe redefining PAGE_MASK so it didn't do the wrong thing on 64 bit types would be better. (That's definitely way out of my depth though.) -- PMM ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: [PATCH] arm: Handle device tree memory regions larger than 4GB
On 07/07/2012 05:36 AM, Peter Maydell wrote: > On 6 July 2012 20:26, Dave Martin wrote: >> On Fri, Jul 06, 2012 at 07:07:35PM +0100, Peter Maydell wrote: >>> On 6 July 2012 00:27, Rob Herring wrote: I would just change arm_add_memory to use phys_addr_t for the size param. This ultimately calls memblock functions which use phys_addr_t for sizes. >>> >>> So I have a patch that does this which basically works. However >>> there is a bit I'm not sure about. arm_add_memory() does this: >>>bank->size = size & PAGE_MASK; >>> >>> in an attempt to clear the bottom bits of the size. However, >>> since PAGE_MASK is defined as: >>> #define PAGE_SIZE (_AC(1,UL) << PAGE_SHIFT) >>> #define PAGE_MASK (~(PAGE_SIZE-1)) >>> >>> PAGE_MASK is a 32 bit unsigned constant and so this & will >>> clear the top 32 bits of bank->size. >>> >>> I'm really not sure what the best way to fix this is; suggestions? >> >> Maybe something like >> >> ~(phys_addr_t)~PAGE_MASK >> >> or >> >> ~(phys_addr_t)(PAGE_SIZE - 1) >> >> would work. > > Mmm. It feels a bit unsatisfactory that an "obviously correct" > line of code like "size &= ~PAGE_MASK" could actually be subtly > wrong (seems like the kind of thing that could easily slip > through code review), so I was wondering if maybe redefining > PAGE_MASK so it didn't do the wrong thing on 64 bit types > would be better. (That's definitely way out of my depth though.) > While there is a mixture of usage of PAGE_MASK in the kernel, I think correct usage is with virt addresses. For phys addresses, there is PHYS_MASK which is sized correctly for LPAE. You really should post this on the lakml list. Rob > -- PMM > ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: Nexus 7/Jellybean Kernel Config
On Wed, Jul 04, 2012 at 10:31:30PM +0200, Marcin Juszkiewicz wrote: > W dniu 04.07.2012 22:19, Christian Robottom Reis pisze: > > Hey there, > > > > I stumbled onto a thread on xdadevelopers where they are looking at > > the Nexus 7; apart from partition and df output, there's a 3.1.10 kernel > > config, which I'm attaching here for convenience and analysis. > > > > ls -l /dev/block/platform/sdhci-tegra.3/by-name/ > .. > > lrwxrwxrwx root root 2012-06-28 11:51 MDA -> /dev/block/mmcblk0p8 > > lrwxrwxrwx root root 2012-06-28 11:51 UDA -> /dev/block/mmcblk0p9 > > Patched kernel to have 9 partitions on mmc... Why is this not upstream? -- Christian Robottom Reis, Engineering VP Brazil (GMT-3) | [+55] 16 9112 6430 | [+1] 612 216 4935 Linaro.org: Open Source Software for ARM SoCs ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
Re: Nexus 7/Jellybean Kernel Config
Shouldn't require a kernel change. Just a configuration with a really annoyingly low default. I am concerned with the explosion of the numbers mmc partitions in android though... On Jul 7, 2012 12:09 PM, "Christian Robottom Reis" wrote: > On Wed, Jul 04, 2012 at 10:31:30PM +0200, Marcin Juszkiewicz wrote: > > W dniu 04.07.2012 22:19, Christian Robottom Reis pisze: > > > Hey there, > > > > > > I stumbled onto a thread on xdadevelopers where they are looking at > > > the Nexus 7; apart from partition and df output, there's a 3.1.10 > kernel > > > config, which I'm attaching here for convenience and analysis. > > > > > > ls -l /dev/block/platform/sdhci-tegra.3/by-name/ > > .. > > > lrwxrwxrwx root root 2012-06-28 11:51 MDA -> /dev/block/mmcblk0p8 > > > lrwxrwxrwx root root 2012-06-28 11:51 UDA -> /dev/block/mmcblk0p9 > > > > Patched kernel to have 9 partitions on mmc... > > Why is this not upstream? > -- > Christian Robottom Reis, Engineering VP > Brazil (GMT-3) | [+55] 16 9112 6430 | [+1] 612 216 4935 > Linaro.org: Open Source Software for ARM SoCs > > ___ > linaro-dev mailing list > linaro-dev@lists.linaro.org > http://lists.linaro.org/mailman/listinfo/linaro-dev > ___ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
