This driver adds a basic console that uses the arm JTAG
DCC to transfer data back and forth. It has support for
ARMv6 and ARMv7.
This console is created under the HVC driver, and should be named
/dev/hvcX (or /dev/hvc0 for example).
Cc: Tony Lindgren <tony@atomide.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Nicolas Pitre <nico@fluxnic.net>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Daniel Walker <dwalker@codeaurora.org>
---
drivers/char/Kconfig | 9 +++
drivers/char/Makefile | 1 +
drivers/char/hvc_dcc.c | 133 ++++++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 143 insertions(+), 0 deletions(-)
create mode 100644 drivers/char/hvc_dcc.c
diff --git a/drivers/char/Kconfig b/drivers/char/Kconfig
index 43d3395..d4a7776 100644
--- a/drivers/char/Kconfig
+++ b/drivers/char/Kconfig
@@ -682,6 +682,15 @@ config HVC_UDBG
select HVC_DRIVER
default n
+config HVC_DCC
+ bool "ARM JTAG DCC console"
+ depends on ARM
+ select HVC_DRIVER
+ help
+ This console uses the JTAG DCC on ARM to create a console under the HVC
+ driver. This console is used through a JTAG only on ARM. If you don't have
+ a JTAG then you probably don't want this option.
+
config VIRTIO_CONSOLE
tristate "Virtio console"
depends on VIRTIO
diff --git a/drivers/char/Makefile b/drivers/char/Makefile
index ba53ec9..fa0b824 100644
--- a/drivers/char/Makefile
+++ b/drivers/char/Makefile
@@ -34,6 +34,7 @@ obj-$(CONFIG_HVC_CONSOLE) += hvc_vio.o hvsi.o
obj-$(CONFIG_HVC_ISERIES) += hvc_iseries.o
obj-$(CONFIG_HVC_RTAS) += hvc_rtas.o
obj-$(CONFIG_HVC_TILE) += hvc_tile.o
+obj-$(CONFIG_HVC_DCC) += hvc_dcc.o
obj-$(CONFIG_HVC_BEAT) += hvc_beat.o
obj-$(CONFIG_HVC_DRIVER) += hvc_console.o
obj-$(CONFIG_HVC_IRQ) += hvc_irq.o
diff --git a/drivers/char/hvc_dcc.c b/drivers/char/hvc_dcc.c
new file mode 100644
index 0000000..6470f63
--- /dev/null
+++ b/drivers/char/hvc_dcc.c
@@ ...Acked-by: Nicolas Pitre <nicolas.pitre@linaro.org> This doesn't support both ARMv6 and ARMv7 at run time, but this can trivially be added later when needed. Nicolas --
I was about to make a similar comment when I saw yours ;-) --
Without marking this asm volatile my compiler decides it can cache the
value of __ret in a register and then check the value of it continually
in hvc_dcc_put_chars() (I had to replace get_wait/put_wait with 1 and
fixup the branch otherwise my disassembler barfed on __dcc_(get|put)char).
00000000 <hvc_dcc_put_chars>:
0: ee103e11 mrc 14, 0, r3, cr0, cr1, {0}
4: e3a0c000 mov ip, #0 ; 0x0
8: e2033202 and r3, r3, #536870912 ; 0x20000000
c: ea000006 b 2c <hvc_dcc_put_chars+0x2c>
10: e3530000 cmp r3, #0 ; 0x0
14: 1afffffd bne 10 <hvc_dcc_put_chars+0x10>
18: e7d1000c ldrb r0, [r1, ip]
1c: ee10fe11 mrc 14, 0, pc, cr0, cr1, {0}
20: 2afffffd bcs 1c <hvc_dcc_put_chars+0x1c>
24: ee000e15 mcr 14, 0, r0, cr0, cr5, {0}
28: e28cc001 add ip, ip, #1 ; 0x1
2c: e15c0002 cmp ip, r2
30: bafffff6 blt 10 <hvc_dcc_put_chars+0x10>
34: e1a00002 mov r0, r2
38: e12fff1e bx lr
As you can see, the value of the mrc is checked against DCC_STATUS_TX
(bit 29) and then stored in r3 for later use. Marking this volatile
produces the following:
00000000 <hvc_dcc_put_chars>:
0: e3a03000 mov r3, #0 ; 0x0
4: ea000007 b 28 <hvc_dcc_put_chars+0x28>
8: ee100e11 mrc 14, 0, r0, cr0, cr1, {0}
c: e3100202 tst r0, #536870912 ; 0x20000000
10: 1afffffc bne 8 <hvc_dcc_put_chars+0x8>
14: e7d10003 ldrb r0, [r1, r3]
18: ee10fe11 mrc 14, 0, pc, cr0, cr1, {0}
1c: 2afffffd bcs 18 <hvc_dcc_put_chars+0x18>
20: ee000e15 mcr 14, 0, r0, cr0, cr5, {0}
24: e2833001 add r3, r3, #1 ; 0x1
28: e1530002 cmp r3, r2
2c: bafffff5 blt 8 <hvc_dcc_put_chars+0x8>
30: e1a00002 mov r0, r2
34: ...Are you talking about __dcc_getstatus only? I don't think adding
We could maybe drop the looping for TX, but RX has no C based looping
even tho for v7 it's recommended that we loop (presumably for v6 it's
This is primarily why these function look they way they do. I'd like to
hear from Tony, because he apparent didn't think they did the same
Like this?
for (i = 0; i < count; ++i) {
if (__dcc_getstatus() & DCC_STATUS_RX)
buf[i] = __dcc_getchar();
else
break;
}
It's a micro clean up I guess ..
Daniel
--
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
--
For this, or any other changes you want, I'll gladly take a follow-on patch as this one is already in my tty-next tree. thanks, greg k-h --
The inline assembly differences for v6 vs. v7 in the hvc_dcc
driver are purely optimizations. On a v7 processor, an mrc with
the pc sets the condition codes to the 28-31 bits of the register
being read. It just so happens that the TX/RX full bits the DCC
driver is testing for are high enough in the register to be put
into the condition codes. On a v6 processor, this "feature" isn't
implemented and thus we have to do the usual read, mask, test
operations to check for TX/RX full.
Since we already test the RX/TX full bits before calling
__dcc_getchar() and __dcc_putchar() we don't actually need to do
anything special for v7 over v6. The only difference is in
hvc_dcc_get_chars(). We would test RX full, poll RX full, and
then read a character from the buffer, whereas now we will test
RX full, read a character from the buffer, and then test RX full
again for the second iteration of the loop. It doesn't seem
possible for the buffer to go from full to empty between testing
the RX full and reading a character. Therefore, replace the v7
versions with the v6 versions and everything works the same.
While we're here, cleanup the for loops a bit and mark the inline
assembly as volatile. Not marking it volatile causes GCC to cache
the results of the status and RX buffer registers causing
lockups.
Cc: Tony Lindgren <tony@atomide.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Nicolas Pitre <nicolas.pitre@linaro.org>
Cc: Daniel Walker <dwalker@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
---
drivers/char/hvc_dcc.c | 43 +++++++------------------------------------
1 files changed, 7 insertions(+), 36 deletions(-)
diff --git a/drivers/char/hvc_dcc.c b/drivers/char/hvc_dcc.c
index 6470f63..435f6fa 100644
--- a/drivers/char/hvc_dcc.c
+++ b/drivers/char/hvc_dcc.c
@@ -33,54 +33,29 @@
static inline u32 __dcc_getstatus(void)
{
u32 __ret;
-
- asm("mrc p14, 0, %0, c0, c1, 0 @ read comms ctrl reg"
+ asm volatile("mrc p14, 0, %0, c0, c1, 0 @ read comms ctrl reg"
...I would expect to see three patches. One that adds volatile, which appears to be a good fix. Another patch that changes the assembly lines, and another that does the clean up. The last two are more controversial ones. Daniel -- Sent by an consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum. --
Ok. I'll send a series later today to give some more time for feedback. -- Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum. --
I think you can do that right away. Splitting the patch will make that feedback easier to provide. Nicolas --
Casting and anding with 0xff is unnecessary in
hvc_dcc_put_chars() since buf is already a char[].
__dcc_get_char() can't return an int less than 0 since it only
returns a char. Simplify the if statement in hvc_dcc_get_chars()
to take this into account.
Cc: Daniel Walker <dwalker@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
---
drivers/char/hvc_dcc.c | 12 ++++--------
1 files changed, 4 insertions(+), 8 deletions(-)
diff --git a/drivers/char/hvc_dcc.c b/drivers/char/hvc_dcc.c
index 155ec10..ad23cc8 100644
--- a/drivers/char/hvc_dcc.c
+++ b/drivers/char/hvc_dcc.c
@@ -89,7 +89,7 @@ static int hvc_dcc_put_chars(uint32_t vt, const char *buf, int count)
while (__dcc_getstatus() & DCC_STATUS_TX)
cpu_relax();
- __dcc_putchar((char)(buf[i] & 0xFF));
+ __dcc_putchar(buf[i]);
}
return count;
@@ -99,15 +99,11 @@ static int hvc_dcc_get_chars(uint32_t vt, char *buf, int count)
{
int i;
- for (i = 0; i < count; ++i) {
- int c = -1;
-
+ for (i = 0; i < count; ++i)
if (__dcc_getstatus() & DCC_STATUS_RX)
- c = __dcc_getchar();
- if (c < 0)
+ buf[i] = __dcc_getchar();
+ else
break;
- buf[i] = c;
- }
return i;
}
--
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
--
Here are the split patches. The first two patches cleanup and fix the hvc_dcc driver for my compiler. The final patch is more controversial, it removes the v6 and v7 differences in this driver. Stephen Boyd (3): hvc_dcc: Fix bad code generation by marking assembly volatile hvc_dcc: Simplify put_chars()/get_chars() loops hvc_dcc: Simplify assembly for v6 and v7 ARM drivers/char/hvc_dcc.c | 43 +++++++------------------------------------ 1 files changed, 7 insertions(+), 36 deletions(-) -- Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum. --
Without marking the asm __dcc_getstatus() volatile my compiler
decides it can cache the value of __ret in a register and then
check the value of it continually in hvc_dcc_put_chars() (I had
to replace get_wait/put_wait with 1 and fixup the branch
otherwise my disassembler barfed on __dcc_(get|put)char).
00000000 <hvc_dcc_put_chars>:
0: ee103e11 mrc 14, 0, r3, cr0, cr1, {0}
4: e3a0c000 mov ip, #0 ; 0x0
8: e2033202 and r3, r3, #536870912 ; 0x20000000
c: ea000006 b 2c <hvc_dcc_put_chars+0x2c>
10: e3530000 cmp r3, #0 ; 0x0
14: 1afffffd bne 10 <hvc_dcc_put_chars+0x10>
18: e7d1000c ldrb r0, [r1, ip]
1c: ee10fe11 mrc 14, 0, pc, cr0, cr1, {0}
20: 2afffffd bcs 1c <hvc_dcc_put_chars+0x1c>
24: ee000e15 mcr 14, 0, r0, cr0, cr5, {0}
28: e28cc001 add ip, ip, #1 ; 0x1
2c: e15c0002 cmp ip, r2
30: bafffff6 blt 10 <hvc_dcc_put_chars+0x10>
34: e1a00002 mov r0, r2
38: e12fff1e bx lr
As you can see, the value of the mrc is checked against
DCC_STATUS_TX (bit 29) and then stored in r3 for later use.
Marking the asm volatile produces the following:
00000000 <hvc_dcc_put_chars>:
0: e3a03000 mov r3, #0 ; 0x0
4: ea000007 b 28 <hvc_dcc_put_chars+0x28>
8: ee100e11 mrc 14, 0, r0, cr0, cr1, {0}
c: e3100202 tst r0, #536870912 ; 0x20000000
10: 1afffffc bne 8 <hvc_dcc_put_chars+0x8>
14: e7d10003 ldrb r0, [r1, r3]
18: ee10fe11 mrc 14, 0, pc, cr0, cr1, {0}
1c: 2afffffd bcs 18 <hvc_dcc_put_chars+0x18>
20: ee000e15 mcr 14, 0, r0, cr0, cr5, {0}
24: e2833001 add r3, r3, #1 ; 0x1
28: e1530002 cmp r3, r2
2c: bafffff5 blt 8 <hvc_dcc_put_chars+0x8>
30: e1a00002 mov r0, ...Is volatile needed here? If __dcc_getstatus() return value is Same here? Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html --
That's not really the issue being fixed. Without the volatile, the compiler is free to cache and reuse a previously loaded status value. It is important that the status be read each time. I don't think there is a way of indicating that assembly needs to happen for each use, but that it is OK to discard if the value isn't used. 'volatile' is a bit overloaded. -- Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum. --
Ok, thanks for explanation. Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html --
Acked-by: Tony Lindgren <tony@atomide.com> --
Hi, What compiler ? That might be a usefull information to know, espectially 5 years from now when tracing code history. There has been similar issue with gcc 4.5 recently. AFAIK, in the same idea, the final change has been to mark the asm volatile. Thanks, - Arnaud --
Sure, we can replace "my compiler" with "my compiler (arm-eabi-gcc (GCC) 4.4.0)". -- Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum. --
Compiler version doesn't matter -- this is a simple correctness issue. If an inline asm statement provides an output value then the compiler is free to cache that value in a register, unless the inline asm is marked so that the value may change from one invocation to another. Also, the compiler is free to eliminate the inline asm statement entirely as well if the output value provided by that inline asm is never used... unless if the inline asm is marked as having side effects. In both cases the volatile qualifier does indicate that the returned value may change (new status flag state) and that the asm code therein has side effects (e.g. pop a character off a FIFO). Sometimes the desired effect can be indicated by clever usage of parameter constraints, but in this case the volatile keyword is most appropriate. Nicolas --
The inline assembly differences for v6 vs. v7 in the hvc_dcc
driver are purely optimizations. On a v7 processor, an mrc with
the pc sets the condition codes to the 28-31 bits of the register
being read. It just so happens that the TX/RX full bits the DCC
driver is testing for are high enough in the register to be put
into the condition codes. On a v6 processor, this "feature" isn't
implemented and thus we have to do the usual read, mask, test
operations to check for TX/RX full.
Since we already test the RX/TX full bits before calling
__dcc_getchar() and __dcc_putchar() we don't actually need to do
anything special for v7 over v6. The only difference is in
hvc_dcc_get_chars(). We would test RX full, poll RX full, and
then read a character from the buffer, whereas now we will test
RX full, read a character from the buffer, and then test RX full
again for the second iteration of the loop. It doesn't seem
possible for the buffer to go from full to empty between testing
the RX full and reading a character. Therefore, replace the v7
versions with the v6 versions and everything works the same.
Cc: Tony Lindgren <tony@atomide.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Nicolas Pitre <nicolas.pitre@linaro.org>
Cc: Daniel Walker <dwalker@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
---
drivers/char/hvc_dcc.c | 24 ------------------------
1 files changed, 0 insertions(+), 24 deletions(-)
diff --git a/drivers/char/hvc_dcc.c b/drivers/char/hvc_dcc.c
index ad23cc8..435f6fa 100644
--- a/drivers/char/hvc_dcc.c
+++ b/drivers/char/hvc_dcc.c
@@ -40,19 +40,6 @@ static inline u32 __dcc_getstatus(void)
}
-#if defined(CONFIG_CPU_V7)
-static inline char __dcc_getchar(void)
-{
- char __c;
-
- asm volatile("get_wait: mrc p14, 0, pc, c0, c1, 0 \n\
- bne get_wait \n\
- mrc p14, 0, %0, c0, c5, 0 @ read comms data reg"
- : "=r" (__c) : : "cc");
-
- return __c;
-}
-#else
static inline char ...Acked-by: Tony Lindgren <tony@atomide.com> --
Just marking __dcc_getstatus volatile gives me
00000038 <hvc_dcc_get_chars>:
38: ee10fe11 mrc 14, 0, pc, cr0, cr1, {0}
3c: 1afffffd bne 38 <hvc_dcc_get_chars>
40: ee103e15 mrc 14, 0, r3, cr0, cr5, {0}
44: e3a00000 mov r0, #0 ; 0x0
48: e6ef3073 uxtb r3, r3
4c: ea000004 b 64 <hvc_dcc_get_chars+0x2c>
50: ee10ce11 mrc 14, 0, ip, cr0, cr1, {0}
54: e31c0101 tst ip, #1073741824 ; 0x40000000
58: 012fff1e bxeq lr
5c: e7c13000 strb r3, [r1, r0]
60: e2800001 add r0, r0, #1 ; 0x1
64: e1500002 cmp r0, r2
68: bafffff8 blt 50 <hvc_dcc_get_chars+0x18>
6c: e12fff1e bx lr
Seems the compiler keeps the value of __dcc_getchar() in r3 for the
duration of the loop. So we need to mark that one volatile too. I don't
think __dcc_putchar() needs to be marked volatile but it probably
Definitely for TX since it seems like a redundant loop, but I agree RX
code has changed. Instead of
If RX buffer full
Poll for RX buffer full
Read character from RX buffer
we would have
If RX buffer full
Read character from RX buffer
which doesn't seem all that different assuming the RX buffer doesn't go
Yes, it's much clearer that way.
--
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
--
Tony, any thoughts? -- Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum. --
