mips: decrease size of au1xxx_dbdma_pm_regs[][]
PATCH 1/1 mips: decrease size of au1xxx_dbdma_pm_regs[][]
Only registers [0-DDMA_CHANNEL_BASE][0-6] are used by the suspend and resume routines.
---
These routines are on the bottom of the file. Only used are registers:
au1xxx_dbdma_pm_regs[0][0-3]
and
au1xxx_dbdma_pm_regs[1-NUM_DBDMA_CHANS][0-6]
Is my patch right, that assumes that the array can be smaller, or should the storage and recovery of other registers be added?
Roel
arch/mips/alchemy/common/dbdma.c +1 -1
@@ -175,7 +175,7 @@ static dbdev_tab_t dbdev_tab[] = {#define DBDEV_TAB_SIZE ARRAY_SIZE(dbdev_tab)#ifdef CONFIG_PM -static u32 au1xxx_dbdma_pm_regs[NUM_DBDMA_CHANS +1 ][8 ]; +static u32 au1xxx_dbdma_pm_regs[NUM_DBDMA_CHANS ][7 ];#endif
Hi Roel,
Quoted 8 lines · Roel Kluin
Only registers [0-DDMA_CHANNEL_BASE][0-6] are used by the suspend and resume routines.
Signed-off-by Roel Kluin <[email protected]>---
These routines are on the bottom of the file. Only used are registers:
Quoted 8 lines
au1xxx_dbdma_pm_regs[0][0-3]
and
au1xxx_dbdma_pm_regs[1-NUM_DBDMA_CHANS][0-6]
Is my patch right, that assumes that the array can be smaller, or should the storage and recovery of other registers be added?
Actually, I think there's a bug in the save/restore functions.
This:
for (i = 1, addr = DDMA_CHANNEL_BASE; i < NUM_DBDMA_CHANS; i++) {
should be changed to
for (i = 1, addr = DDMA_CHANNEL_BASE; i <= NUM_DBDMA_CHANS; i++)
as there are 16 individual channels (NUM_DBDMA_CHANS) to save/restore plus the global ddma block config (the +1).
And looking closer at it, the last register in a channel can be skipped since it's
read-only (at offset 0x18).
So, 6 slots per channel for 16 channels plus one row for the ddma global config are enough:
static u32 au1xxx_dbdma_pm_regs[NUM_DBDMA_CHANS + 1][6];
Thanks!
Manuel Lauss
Hi Roel,
Quoted 8 lines · Roel Kluin
Only registers [0-DDMA_CHANNEL_BASE][0-6] are used by the suspend and resume routines.
Signed-off-by Roel Kluin <[email protected]>---
These routines are on the bottom of the file. Only used are registers:
Quoted 8 lines
au1xxx_dbdma_pm_regs[0][0-3]
and
au1xxx_dbdma_pm_regs[1-NUM_DBDMA_CHANS][0-6]
Is my patch right, that assumes that the array can be smaller, or should the storage and recovery of other registers be added?
Actually, I think there's a bug in the save/restore functions.
This:
for (i = 1, addr = DDMA_CHANNEL_BASE; i < NUM_DBDMA_CHANS; i++) {
should be changed to
for (i = 1, addr = DDMA_CHANNEL_BASE; i <= NUM_DBDMA_CHANS; i++)
as there are 16 individual channels (NUM_DBDMA_CHANS) to save/restore plus the global ddma block config (the +1).
And looking closer at it, the last register in a channel can be skipped since it's
read-only (at offset 0x18).
So, 6 slots per channel for 16 channels plus one row for the ddma global config are enough:
static u32 au1xxx_dbdma_pm_regs[NUM_DBDMA_CHANS + 1][6];
Thanks!
Manuel Lauss