ARM: LL_DEBUG: Fix hang on i.MX53 when wrong port selected.
Subsystems (2)
RFC PATCH 1/1 ARM: LL_DEBUG: Fix hang on i.MX53 when wrong port selected.
Attempts to access the i.MX53 UART data registers while the UART is not enabled cause the processor to hang.
This can occur when CONFIG_LL_DEBUG is selected but CONFIG_DEBUG_IMX_UART_PORT does not correspond to the port enabled by the bootloader.
So, do not attempt to send any data if the port is not enabled.
Checking the enabled bit requires an extra scratch register. For the moment I have only added this to the imx version. Obviously, if this solution is acceptable, the register would have to be added to all the other implementations too in order to avoid breaking the build.
Hence RFC status for the moment to see if there are any better ideas.
--- arch/arm/boot/compressed/debug.S | 2 +- arch/arm/boot/compressed/head.S | 16 ++++++++-------- arch/arm/include/debug/imx.S | 10 ++++++++-- arch/arm/kernel/debug.S | 2 +- 4 files changed, 18 insertions(+), 12 deletions(-)
arch/arm/boot/compressed/debug.S +1 -1
@@ -8,7 +8,7 @@ ENTRY(putc) addruart r1, r2, r3 waituart r3, r1 - senduart r0, r1 + senduart r0, r1, r3 busyuart r3, r1 mov pc, lr ENDPROC(putc)
arch/arm/boot/compressed/head.S +8 -8
@@ -26,19 +26,19 @@ #if defined(CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K) || defined(CONFIG_CPU_V7) .macro loadsp, rb, tmp .endm - .macro writeb, ch, rb + .macro writeb, ch, rb, tmp mcr p14, 0, \ch, c0, c5, 0 .endm #elif defined(CONFIG_CPU_XSCALE) .macro loadsp, rb, tmp .endm - .macro writeb, ch, rb + .macro writeb, ch, rb, tmp mcr p14, 0, \ch, c8, c0, 0 .endm #else .macro loadsp, rb, tmp .endm - .macro writeb, ch, rb + .macro writeb, ch, rb, tmp mcr p14, 0, \ch, c1, c0, 0 .endm #endif
@@ -47,8 +47,8 @@ #include CONFIG_DEBUG_LL_INCLUDE - .macro writeb, ch, rb - senduart \ch, \rb + .macro writeb, ch, rb, tmp + senduart \ch, \rb, \tmp .endm #if defined(CONFIG_ARCH_SA1100)
@@ -870,7 +870,7 @@ proc_types: W(b) __armv3_mpu_cache_on W(b) __armv3_mpu_cache_off W(b) __armv3_mpu_cache_flush - + .word 0x41009400 @ ARM94x .word 0xff00ff00 W(b) __armv4_mpu_cache_on
@@ -1084,7 +1084,7 @@ __armv4_mpu_cache_flush: mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache mcr p15, 0, ip, c7, c10, 4 @ drain WB mov pc, lr - + __fa526_cache_flush: tst r4, #1 movne pc, lr
@@ -1246,7 +1246,7 @@ puts: loadsp r3, r1 1: ldrb r2, [r0], #1 teq r2, #0 moveq pc, lr -2: writeb r2, r3 +2: writeb r2, r3, r1 mov r1, #0x00020000 3: subs r1, r1, #1 bne 3b
arch/arm/include/debug/imx.S +8 -2
@@ -33,8 +33,14 @@ ldr \rv, =UART_VADDR @ virtual .endm - .macro senduart,rd,rx - str \rd, [\rx, #0x40] @ TXDATA + .macro senduart,rd,rx, tmp +/* + * Data registers only accessible if uart enabled. + * Check to prevent hangs if bootloader config does not match + */ + ldr \tmp, [\rx, #0x80] @ CR1 + tst \tmp, #1 + strne \rd, [\rx, #0x40] @ TXDATA .endm .macro waituart,rd,rx
arch/arm/kernel/debug.S +1 -1
@@ -81,7 +81,7 @@ ENTRY(printascii) addruart_current r3, r1, r2 b 2f 1: waituart r2, r3 - senduart r1, r3 + senduart r1, r3, r2 busyuart r2, r3 teq r1, #'\n' moveq r1, #'\r'
RFC PATCH 1/1 ARM: LL_DEBUG: Fix hang on i.MX53 when wrong port selected.
Quoted 1 line · Martin Fuzzey
Hence RFC status for the moment to see if there are any better ideas.
How about detecting the condition in the addruart macro - if the UART is not enabled, we can clear r1 - and detect that in the various users and just bail out. We wouldn't need to test on every character then.
Signature · 2 lines
FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up according to speedtest.net.
RFC PATCH 1/1 ARM: LL_DEBUG: Fix hang on i.MX53 when wrong port selected.
In reply to Martin FuzzeyHi Martin,
Quoted 9 lines · Martin Fuzzey
Attempts to access the i.MX53 UART data registers while the UART is not enabled cause the processor to hang.
This can occur when
CONFIG_LL_DEBUGis selected butCONFIG_DEBUG_IMX_UART_PORTdoes not correspond to the port enabled by the bootloader.So, do not attempt to send any data if the port is not enabled.
Quoted 8 lines
Checking the enabled bit requires an extra scratch register. For the moment I have only added this to the imx version. Obviously, if this solution is acceptable, the register would have to be added to all the other implementations too in order to avoid breaking the build.
Hence RFC status for the moment to see if there are any better ideas.
I think the current consensus is to not try to fix CONFIG_LL_DEBUG too much as if you use it it means you know what you are doing, refer to the following discussion:
http://lists.infradead.org/pipermail/linux-arm-kernel/2015-January/319187.html
Signature · 3 lines
Alexandre Belloni, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com