| Message ID | 1496825972-26829-1-git-send-email-xieyisheng1@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Jun 07, 2017 at 04:59:32PM +0800, Yisheng Xie wrote:
> This reverts commit 63ce446c9b5787a94ed875bab20772e1a2b3092f.
There's no real reason to revert this, thereby removing it from non-LPAE
cases where it does work. You've already shown adding a dependency for
LPAE on this option (or even vice versa) is enough to stop the problem
occuring.
IMHO, It would be better to make VMSPLIT_3G_OPT depend on !ARM_LPAE.
Hi Russell, Thanks for comment! On 2017/6/7 17:33, Russell King - ARM Linux wrote: > On Wed, Jun 07, 2017 at 04:59:32PM +0800, Yisheng Xie wrote: >> This reverts commit 63ce446c9b5787a94ed875bab20772e1a2b3092f. > > There's no real reason to revert this, thereby removing it from non-LPAE > cases where it does work. You've already shown adding a dependency for > LPAE on this option (or even vice versa) is enough to stop the problem > occuring. > > IMHO, It would be better to make VMSPLIT_3G_OPT depend on !ARM_LPAE. OK, I will send a patch as your suggestion. Thanks Yisheng Xie
diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 4c1a35f..5961835a 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -1415,8 +1415,6 @@ choice config VMSPLIT_3G bool "3G/1G user/kernel split" - config VMSPLIT_3G_OPT - bool "3G/1G user/kernel split (for full 1G low memory)" config VMSPLIT_2G bool "2G/2G user/kernel split" config VMSPLIT_1G @@ -1428,7 +1426,6 @@ config PAGE_OFFSET default PHYS_OFFSET if !MMU default 0x40000000 if VMSPLIT_1G default 0x80000000 if VMSPLIT_2G - default 0xB0000000 if VMSPLIT_3G_OPT default 0xC0000000 config NR_CPUS
This reverts commit 63ce446c9b5787a94ed875bab20772e1a2b3092f. When enable VMSPLIT_3G_OPT, the PAGE_OFFSET will be 0xB0000000, and then we will get the value "1" for TTBR1_SIZE(T1SZ) as the following MACRO: #define TTBR1_SIZE (((PAGE_OFFSET >> 30) - 1) << 16) which means TTBR1's range will be 0x80000000 to 0xffffffff, and it is waster of VA [0x80000000, 0xB0000000). So it is better to use VMSPLIT_2G instead, to achieve the goal "allows for 1GB systems to have their RAM entirely mapped as low memory". Meanwhile, It will trigger a bug of boot fail with CONFIG_ARM_LPAE=y, which I reported: https://patchwork.kernel.org/patch/9768573/ For ARM_LPAE only support 3:1, 2:2, 1:3 split of TTBR1, which mention in: http://elinux.org/images/6/6a/Elce11_marinas.pdf - p16 Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com> --- arch/arm/Kconfig | 3 --- 1 file changed, 3 deletions(-)