[PATCH] arm: LPAE: kexec: clear TTBCR.T1SZ explicitly when PHYS_OFFSET > PAGE_OFFSET

 Clearing TTBCR.T1SZ explicitly when kernel runs on a configuration of
PHYS_OFFSET > PAGE_OFFSET.
 Reading TTBCR in early boot stage might returns the value of the
previous kernel's configuration, especially in case of kexec. For
example, if normal kernel (first kernel) had run on a configuration
of PHYS_OFFSET <= PAGE_OFFSET and crash kernel (second kernel) is
running on a configuration PHYS_OFFSET > PAGE_OFFSET, which can happen
because it depends on the reserved area for crash kernel, reading
TTBCR and using the value without clearing TTBCR.T1SZ might risky
because the value doesn't have a reset value for TTBCR.T1SZ.

Signed-off-by: Hoeun Ryu <redacted>
---
 arch/arm/mm/proc-v7-3level.S | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm/mm/proc-v7-3level.S b/arch/arm/mm/proc-v7-3level.S
index 5e5720e..9ac2bec 100644
--- a/arch/arm/mm/proc-v7-3level.S
+++ b/arch/arm/mm/proc-v7-3level.S

@@ -140,6 +140,7 @@ ENDPROC(cpu_v7_set_pte_ext)
 	 * otherwise booting secondary CPUs would end up using TTBR1 for the
 	 * identity mapping set up in TTBR0.
 	 */
+	bichi	\tmp, \tmp, #(1 << 16)				@ clear TTBCR.T1SZ
 	orrls	\tmp, \tmp, #TTBR1_SIZE				@ TTBCR.T1SZ
 	mcr	p15, 0, \tmp, c2, c0, 2				@ TTBCR
 	mov	\tmp, \ttbr1, lsr #20

-- 
2.7.4

On Mon, Jun 05, 2017 at 06:22:20PM +0900, Hoeun Ryu wrote:

This looks insufficient.  There's two bits here:

 * TTBR0/TTBR1 split (PAGE_OFFSET):
 *   0x40000000: T0SZ = 2, T1SZ = 0 (not used)
 *   0x80000000: T0SZ = 0, T1SZ = 1
 *   0xc0000000: T0SZ = 0, T1SZ = 2

but you seem to only be clearing one bit.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently@9.6Mbps down 400kbps up
according to speedtest.net.

On Mon, Jun 5, 2017 at 6:34 PM, Russell King - ARM Linux
[off-list ref] wrote:

On Mon, Jun 05, 2017 at 06:22:20PM +0900, Hoeun Ryu wrote:

quoted

diff --git a/arch/arm/mm/proc-v7-3level.S b/arch/arm/mm/proc-v7-3level.S
index 5e5720e..9ac2bec 100644
--- a/arch/arm/mm/proc-v7-3level.S
+++ b/arch/arm/mm/proc-v7-3level.S

@@ -140,6 +140,7 @@ ENDPROC(cpu_v7_set_pte_ext)
       * otherwise booting secondary CPUs would end up using TTBR1 for the
       * identity mapping set up in TTBR0.
       */
+     bichi   \tmp, \tmp, #(1 << 16)                          @ clear TTBCR.T1SZ

This looks insufficient.  There's two bits here:

 * TTBR0/TTBR1 split (PAGE_OFFSET):
 *   0x40000000: T0SZ = 2, T1SZ = 0 (not used)
 *   0x80000000: T0SZ = 0, T1SZ = 1
 *   0xc0000000: T0SZ = 0, T1SZ = 2

but you seem to only be clearing one bit.

Oh, I'm sorry for the mistake, I'll fix this like #(7 << 16) in v2.
(There're 3 bits in TxSZ)
Thank you for the review.

--
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.