ARM: tegra: use setup_mm_for_reboot rather than explicit pgd switch
Subsystems (3)
PATCH 1/2 ARM: tegra: use setup_mm_for_reboot rather than explicit pgd switch
This patch changes the Tegra PM code to use the setup_mm_for_reboot helper rather than call cpu_switch_mm directly. This keeps things like TLB invalidation in one place.
--- arch/arm/mach-tegra/pm.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-)
arch/arm/mach-tegra/pm.c +1 -6
@@ -183,12 +183,7 @@ bool tegra_set_cpu_in_lp2(int phy_cpu_id)static int tegra_sleep_cpu (unsigned long v2p) { -/* Switch to the identity mapping. */ -cpu_switch_mm (idmap_pgd, &init_mm); - -/* Flush the TLB. */ -local_flush_tlb_all (); - +setup_mm_for_reboot ();tegra_sleep_cpu_finish (v2p);/* should never here */
Signature · 1 line
1.8.0
PATCH 2/2 ARM: tegra: use CS_LAR_KEY definition instead of 0xC5ACCE55
When enabling coresight on secondary cores for tegra, make use of our fancy new definition for the magic lock code.
--- arch/arm/mach-tegra/headsmp.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/mach-tegra/headsmp.S +2 -1
@@ -1,6 +1,7 @@ #include <linux/linkage.h> #include <linux/init.h> +#include <asm/hardware/coresight.h> #include "sleep.h" .section ".text.head", "ax"
@@ -8,7 +9,7 @@ ENTRY(tegra_secondary_startup) bl v7_invalidate_l1 /* Enable coresight */ - mov32 r0, 0xC5ACCE55 + mov32 r0, CS_LAR_KEY mcr p14, 0, r0, c7, c12, 6 b secondary_startup ENDPROC(tegra_secondary_startup)
Signature · 1 line
1.8.0
PATCH 2/2 ARM: tegra: use CS_LAR_KEY definition instead of 0xC5ACCE55
Quoted 2 lines · Will Deacon
When enabling coresight on secondary cores for tegra, make use of our fancy new definition for the magic lock code.
The code you're editing has actually been removed in the 3.10 cycle, so there's no need for this patch any more.
Stephen Warren
AppliedPATCH 1/2 ARM: tegra: use setup_mm_for_reboot rather than explicit pgd switch
In reply to Will DeaconQuoted 3 lines · Will Deacon
This patch changes the Tegra PM code to use the
setup_mm_for_reboothelper rather than callcpu_switch_mmdirectly. This keeps things like TLB invalidation in one place.
I've applied this to Tegra's for-3.10/cleanup branch.