[U-Boot] [PATCH 00/16] image: Fix various test failures
Recent changes have broken the FIT and vboot tests. Also the SPI tests have been wrong since before the last release and were disabled.
This series collects together the required fixes.
Note: The FIT and vboot tests are hard to run (in that each requires manual effort). At some point we should be able to bring these into Stephen Warren's test framework.
Simon Glass (16): image: Correct the OS location code to work on sandbox Revert "image-fit: Fix signature checking" image: Fix FIT and vboot tests to exit sandbox correctly trace: Fix compiler warnings in trace lib: Don't instrument the div64 function trace: Improve the trace test number recognition timer: Support tracing fully timer: Provide an early timer timer: Set up the real timer after driver model is available sandbox: timer: Support the early timer sandbox: Correct ordering of defconfig sandbox: Enable the early timer sandbox: spi: Add more debugging to SPI emulation sandbox: spi: Remove an incorrect free() spi: Correct two error return values spi: Re-enable the SPI flash tests
cmd/trace.c | 4 ++-- common/board_f.c | 6 ++++++ common/board_r.c | 14 ++++++++++++-- common/bootm.c | 2 +- common/image-fit.c | 16 +++++++++++++--- configs/sandbox_defconfig | 11 ++++++----- drivers/mtd/spi/sandbox.c | 14 ++++++++++---- drivers/mtd/spi/sf_probe.c | 4 +--- drivers/mtd/spi/spi_flash.c | 2 +- drivers/timer/Kconfig | 10 ++++++++++ drivers/timer/sandbox_timer.c | 14 ++++++++++++-- drivers/timer/timer-uclass.c | 6 +++--- include/image.h | 5 +---- include/timer.h | 21 +++++++++++++++++++++ lib/div64.c | 3 ++- lib/time.c | 28 +++++++++++++++++++++------- test/dm/Makefile | 4 ++-- test/image/test-fit.py | 4 ++++ test/trace/test-trace.sh | 4 +++- test/vboot/sandbox-u-boot.dts | 3 +++ 20 files changed, 134 insertions(+), 41 deletions(-)
A recent change broke the 'bootm' command on sandbox. The root cause is using a pointer as an address. Conversion from pointer to address needs to use map_to_sysmem() so that sandbox can do the right thing. The problem was pre-existing but uncovered by a recent commit.
Fix this. Also move fit_get_end() to the C file to avoid needing to include mapmem.h (and thus asm/io.h) everywhere.
Fixes: 1fec3c5d (common/image.c: Make boot_get_ramdisk() perform a check for Android images)
Signed-off-by: Simon Glass sjg@chromium.org ---
common/bootm.c | 2 +- common/image-fit.c | 5 +++++ include/image.h | 5 +---- 3 files changed, 7 insertions(+), 5 deletions(-)
diff --git a/common/bootm.c b/common/bootm.c index 99d574d..df27089 100644 --- a/common/bootm.c +++ b/common/bootm.c @@ -201,7 +201,7 @@ static int bootm_find_os(cmd_tbl_t *cmdtp, int flag, int argc, images.ep += images.os.load; }
- images.os.start = (ulong)os_hdr; + images.os.start = map_to_sysmem(os_hdr);
return 0; } diff --git a/common/image-fit.c b/common/image-fit.c index c531ee7..d3fad30 100644 --- a/common/image-fit.c +++ b/common/image-fit.c @@ -851,6 +851,11 @@ static int fit_image_hash_get_ignore(const void *fit, int noffset, int *ignore) return 0; }
+ulong fit_get_end(const void *fit) +{ + return map_to_sysmem((void *)(fit + fdt_totalsize(fit))); +} + /** * fit_set_timestamp - set node timestamp property * @fit: pointer to the FIT format image header diff --git a/include/image.h b/include/image.h index 299d6d2..518a4f5 100644 --- a/include/image.h +++ b/include/image.h @@ -818,10 +818,7 @@ static inline ulong fit_get_size(const void *fit) * returns: * end address of the FIT image (blob) in memory */ -static inline ulong fit_get_end(const void *fit) -{ - return (ulong)fit + fdt_totalsize(fit); -} +ulong fit_get_end(const void *fit);
/** * fit_get_name - get FIT node name
This reverts commit 84ca65aa4bd0d03867e9e49805201d0564d3ffb0.
On signature verification failures fit_image_verify() should NOT exit with error. Only keys marked 'required' can cause image verification failure. This logic is already there and works correctly.
Add a comment to make this clear.
Fixes: 84ca65aa (image-fit: Fix signature checking) Signed-off-by: Simon Glass sjg@chromium.org ---
common/image-fit.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
diff --git a/common/image-fit.c b/common/image-fit.c index d3fad30..fbd9e0d 100644 --- a/common/image-fit.c +++ b/common/image-fit.c @@ -1035,10 +1035,15 @@ int fit_image_verify(const void *fit, int image_noffset) strlen(FIT_SIG_NODENAME))) { ret = fit_image_check_sig(fit, noffset, data, size, -1, &err_msg); - if (ret) { + + /* + * Show an indication on failure, but do not return + * an error. Only keys marked 'required' can cause + * an image validation failure. See the call to + * fit_image_verify_required_sigs() above. + */ + if (ret) puts("- "); - goto error; - } else puts("+ "); }
When used with a device tree, sandbox now requires a 'reset' controller. Add this to the device trees so that reset works and the tests can complete.
Signed-off-by: Simon Glass sjg@chromium.org Fixes: 5010d98f (sandbox: Use the reset driver to handle reset) ---
test/image/test-fit.py | 4 ++++ test/vboot/sandbox-u-boot.dts | 3 +++ 2 files changed, 7 insertions(+)
diff --git a/test/image/test-fit.py b/test/image/test-fit.py index d5143cb..db0649f 100755 --- a/test/image/test-fit.py +++ b/test/image/test-fit.py @@ -108,6 +108,10 @@ base_fdt = ''' model = "Sandbox Verified Boot Test"; compatible = "sandbox";
+ reset@0 { + compatible = "sandbox,reset"; + }; + }; '''
diff --git a/test/vboot/sandbox-u-boot.dts b/test/vboot/sandbox-u-boot.dts index a1e853c..63f8f40 100644 --- a/test/vboot/sandbox-u-boot.dts +++ b/test/vboot/sandbox-u-boot.dts @@ -4,4 +4,7 @@ model = "Sandbox Verified Boot Test"; compatible = "sandbox";
+ reset@0 { + compatible = "sandbox,reset"; + }; };
With min() we must use the same type for each parameter. Fix two problems in trace.c which produce compiler warnings.
Signed-off-by: Simon Glass sjg@chromium.org ---
cmd/trace.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/cmd/trace.c b/cmd/trace.c index 1e62a1a..1a6d8c3 100644 --- a/cmd/trace.c +++ b/cmd/trace.c @@ -43,7 +43,7 @@ static int create_func_list(int argc, char * const argv[]) err = trace_list_functions(buff + buff_ptr, avail, &needed); if (err) printf("Error: truncated (%#x bytes needed)\n", needed); - used = min(avail, needed); + used = min(avail, (size_t)needed); printf("Function trace dumped to %08lx, size %#zx\n", (ulong)map_to_sysmem(buff + buff_ptr), used); setenv_hex("profbase", map_to_sysmem(buff)); @@ -67,7 +67,7 @@ static int create_call_list(int argc, char * const argv[]) err = trace_list_calls(buff + buff_ptr, avail, &needed); if (err) printf("Error: truncated (%#x bytes needed)\n", needed); - used = min(avail, needed); + used = min(avail, (size_t)needed); printf("Call list dumped to %08lx, size %#zx\n", (ulong)map_to_sysmem(buff + buff_ptr), used);
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
With min() we must use the same type for each parameter. Fix two problems in trace.c which produce compiler warnings.
Signed-off-by: Simon Glass sjg@chromium.org
cmd/trace.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
This function can be called from the timer code on instrumented functions. Mark it as 'notrace' so that it doesn't cause infinite recursion.
Signed-off-by: Simon Glass sjg@chromium.org ---
lib/div64.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/div64.c b/lib/div64.c index 795ef0e..319fca5 100644 --- a/lib/div64.c +++ b/lib/div64.c @@ -18,8 +18,9 @@
#include <div64.h> #include <linux/types.h> +#include <linux/compiler.h>
-uint32_t __div64_32(uint64_t *n, uint32_t base) +uint32_t notrace __div64_32(uint64_t *n, uint32_t base) { uint64_t rem = *n; uint64_t b = base;
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
This function can be called from the timer code on instrumented functions. Mark it as 'notrace' so that it doesn't cause infinite recursion.
Signed-off-by: Simon Glass sjg@chromium.org
lib/div64.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
The awk tool can be confused by return character (ASCII 13) in its input since it thinks there is a separate field. These can appear if the terminal is in raw mode, perhaps due to a previous U-Boot crash with sandbox. This is very confusing. Remove these so that the trace test passes.
Signed-off-by: Simon Glass sjg@chromium.org ---
test/trace/test-trace.sh | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/test/trace/test-trace.sh b/test/trace/test-trace.sh index 3e8651e..746793c 100755 --- a/test/trace/test-trace.sh +++ b/test/trace/test-trace.sh @@ -45,7 +45,9 @@ check_results() { # between calls 2 and 3, where tracing is paused. # This code gets the sign of the difference between each number and # its predecessor. - counts="$(tr -d , <${tmp} | awk '/traced function calls/ { diff = $1 - upto; upto = $1; printf "%d ", diff < 0 ? -1 : (diff > 0 ? 1 : 0)}')" + counts="$(tr -d ',\r' <${tmp} | awk \ + '/traced function calls/ { diff = $1 - upto; upto = $1; \ + printf "%d ", diff < 0 ? -1 : (diff > 0 ? 1 : 0)}')"
if [ "${counts}" != "1 1 0 1 " ]; then fail "trace collection error: ${counts}"
A few of the functions in the timer uclass are not marked with 'notrace'. Fix this so that tracing can be used with CONFIG_TRACE.
Signed-off-by: Simon Glass sjg@chromium.org ---
drivers/timer/timer-uclass.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/timer/timer-uclass.c b/drivers/timer/timer-uclass.c index 83d1a35..382c0f2 100644 --- a/drivers/timer/timer-uclass.c +++ b/drivers/timer/timer-uclass.c @@ -22,7 +22,7 @@ DECLARE_GLOBAL_DATA_PTR; * tick, and no timer interrupt. */
-int timer_get_count(struct udevice *dev, u64 *count) +int notrace timer_get_count(struct udevice *dev, u64 *count) { const struct timer_ops *ops = device_get_ops(dev);
@@ -32,9 +32,9 @@ int timer_get_count(struct udevice *dev, u64 *count) return ops->get_count(dev, count); }
-unsigned long timer_get_rate(struct udevice *dev) +unsigned long notrace timer_get_rate(struct udevice *dev) { - struct timer_dev_priv *uc_priv = dev_get_uclass_priv(dev); + struct timer_dev_priv *uc_priv = dev->uclass_priv;
return uc_priv->clock_rate; }
Hi Simon,
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
A few of the functions in the timer uclass are not marked with 'notrace'. Fix this so that tracing can be used with CONFIG_TRACE.
Signed-off-by: Simon Glass sjg@chromium.org
drivers/timer/timer-uclass.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/timer/timer-uclass.c b/drivers/timer/timer-uclass.c index 83d1a35..382c0f2 100644 --- a/drivers/timer/timer-uclass.c +++ b/drivers/timer/timer-uclass.c @@ -22,7 +22,7 @@ DECLARE_GLOBAL_DATA_PTR;
- tick, and no timer interrupt.
*/
-int timer_get_count(struct udevice *dev, u64 *count) +int notrace timer_get_count(struct udevice *dev, u64 *count) { const struct timer_ops *ops = device_get_ops(dev);
@@ -32,9 +32,9 @@ int timer_get_count(struct udevice *dev, u64 *count) return ops->get_count(dev, count);
Besides making timer_get_count() and timer_get_rate() APIs notrace, does it help to make the timer uclass ops notrace as well?
}
-unsigned long timer_get_rate(struct udevice *dev) +unsigned long notrace timer_get_rate(struct udevice *dev) {
struct timer_dev_priv *uc_priv = dev_get_uclass_priv(dev);
struct timer_dev_priv *uc_priv = dev->uclass_priv;
Why is this change needed?
return uc_priv->clock_rate;}
Regards, Bin
Hi Bin,
On 16 February 2016 at 02:21, Bin Meng bmeng.cn@gmail.com wrote:
Hi Simon,
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
A few of the functions in the timer uclass are not marked with 'notrace'. Fix this so that tracing can be used with CONFIG_TRACE.
Signed-off-by: Simon Glass sjg@chromium.org
drivers/timer/timer-uclass.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/timer/timer-uclass.c b/drivers/timer/timer-uclass.c index 83d1a35..382c0f2 100644 --- a/drivers/timer/timer-uclass.c +++ b/drivers/timer/timer-uclass.c @@ -22,7 +22,7 @@ DECLARE_GLOBAL_DATA_PTR;
- tick, and no timer interrupt.
*/
-int timer_get_count(struct udevice *dev, u64 *count) +int notrace timer_get_count(struct udevice *dev, u64 *count) { const struct timer_ops *ops = device_get_ops(dev);
@@ -32,9 +32,9 @@ int timer_get_count(struct udevice *dev, u64 *count) return ops->get_count(dev, count);
Besides making timer_get_count() and timer_get_rate() APIs notrace, does it help to make the timer uclass ops notrace as well?
I don't think so. From what I can this needs to be done in each driver.
}
-unsigned long timer_get_rate(struct udevice *dev) +unsigned long notrace timer_get_rate(struct udevice *dev) {
struct timer_dev_priv *uc_priv = dev_get_uclass_priv(dev);
struct timer_dev_priv *uc_priv = dev->uclass_priv;Why is this change needed?
To avoid a function call, and the notrace problem.
return uc_priv->clock_rate;}
Regards, Bin
Regards, Simon
In some cases the timer must be accessible before driver model is active. Examples include when using CONFIG_TRACE to trace U-Boot's execution before driver model is set up. Enable this option to use an early timer. These functions must be supported by your timer driver: timer_early_get_count() and timer_early_get_rate().
Signed-off-by: Simon Glass sjg@chromium.org ---
drivers/timer/Kconfig | 10 ++++++++++ include/timer.h | 21 +++++++++++++++++++++ lib/time.c | 28 +++++++++++++++++++++------- 3 files changed, 52 insertions(+), 7 deletions(-)
diff --git a/drivers/timer/Kconfig b/drivers/timer/Kconfig index ff65a73..cb18f12 100644 --- a/drivers/timer/Kconfig +++ b/drivers/timer/Kconfig @@ -9,6 +9,16 @@ config TIMER will be used. The timer is usually a 32 bits free-running up counter. There may be no real tick, and no timer interrupt.
+config TIMER_EARLY + bool "Allow timer to be used early in U-Boot" + depends on TIMER + help + In some cases the timer must be accessible before driver model is + active. Examples include when using CONFIG_TRACE to trace U-Boot's + execution before driver model is set up. Enable this option to + use an early timer. These functions must be supported by your timer + driver: timer_early_get_count() and timer_early_get_rate(). + config ALTERA_TIMER bool "Altera timer support" depends on TIMER diff --git a/include/timer.h b/include/timer.h index f14725c..a503bfd 100644 --- a/include/timer.h +++ b/include/timer.h @@ -67,4 +67,25 @@ struct timer_dev_priv { unsigned long clock_rate; };
+/** + * timer_early_get_count() - Implement timer_get_count() before driver model + * + * If CONFIG_TIMER_EARLY is enabled, this function wil be called to return + * the current timer value before the proper driver model timer is ready. + * It should be implemented by one of the timer values. This is mostly useful + * for tracing. + */ +u64 timer_early_get_count(void); + +/** + * timer_early_get_rate() - Get the timer rate before driver model + * + * If CONFIG_TIMER_EARLY is enabled, this function wil be called to return + * the current timer rate in Hz before the proper driver model timer is ready. + * It should be implemented by one of the timer values. This is mostly useful + * for tracing. This corresponds to the clock_rate value in struct + * timer_dev_priv. + */ +unsigned long timer_early_get_rate(void); + #endif /* _TIMER_H_ */ diff --git a/lib/time.c b/lib/time.c index e9f6861..f37150f 100644 --- a/lib/time.c +++ b/lib/time.c @@ -43,11 +43,17 @@ extern unsigned long __weak timer_read_counter(void); #ifdef CONFIG_TIMER ulong notrace get_tbclk(void) { - int ret; + if (!gd->timer) { +#ifdef CONFIG_TIMER_EARLY + return timer_early_get_rate(); +#else + int ret;
- ret = dm_timer_init(); - if (ret) - return ret; + ret = dm_timer_init(); + if (ret) + return ret; +#endif + }
return timer_get_rate(gd->timer); } @@ -57,9 +63,17 @@ uint64_t notrace get_ticks(void) u64 count; int ret;
- ret = dm_timer_init(); - if (ret) - return ret; + if (!gd->timer) { +#ifdef CONFIG_TIMER_EARLY + return timer_early_get_count(); +#else + int ret; + + ret = dm_timer_init(); + if (ret) + return ret; +#endif + }
ret = timer_get_count(gd->timer, &count); if (ret)
Hi Simon,
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
In some cases the timer must be accessible before driver model is active. Examples include when using CONFIG_TRACE to trace U-Boot's execution before driver model is set up. Enable this option to use an early timer. These functions must be supported by your timer driver: timer_early_get_count() and timer_early_get_rate().
Signed-off-by: Simon Glass sjg@chromium.org
drivers/timer/Kconfig | 10 ++++++++++ include/timer.h | 21 +++++++++++++++++++++ lib/time.c | 28 +++++++++++++++++++++------- 3 files changed, 52 insertions(+), 7 deletions(-)
diff --git a/drivers/timer/Kconfig b/drivers/timer/Kconfig index ff65a73..cb18f12 100644 --- a/drivers/timer/Kconfig +++ b/drivers/timer/Kconfig @@ -9,6 +9,16 @@ config TIMER will be used. The timer is usually a 32 bits free-running up counter. There may be no real tick, and no timer interrupt.
+config TIMER_EARLY
bool "Allow timer to be used early in U-Boot"depends on TIMERhelpIn some cases the timer must be accessible before driver model isactive. Examples include when using CONFIG_TRACE to trace U-Boot'sexecution before driver model is set up. Enable this option touse an early timer. These functions must be supported by your timerdriver: timer_early_get_count() and timer_early_get_rate().config ALTERA_TIMER bool "Altera timer support" depends on TIMER diff --git a/include/timer.h b/include/timer.h index f14725c..a503bfd 100644 --- a/include/timer.h +++ b/include/timer.h @@ -67,4 +67,25 @@ struct timer_dev_priv { unsigned long clock_rate; };
+/**
- timer_early_get_count() - Implement timer_get_count() before driver model
- If CONFIG_TIMER_EARLY is enabled, this function wil be called to return
- the current timer value before the proper driver model timer is ready.
- It should be implemented by one of the timer values. This is mostly useful
- for tracing.
- */
+u64 timer_early_get_count(void);
+/**
- timer_early_get_rate() - Get the timer rate before driver model
- If CONFIG_TIMER_EARLY is enabled, this function wil be called to return
- the current timer rate in Hz before the proper driver model timer is ready.
nits: two spaces between 'rate' and 'in'
- It should be implemented by one of the timer values. This is mostly useful
- for tracing. This corresponds to the clock_rate value in struct
- timer_dev_priv.
Is this supposed to be a hard-codeded value returned by the timer driver? The timer-uclass driver gets this clock rate from device tree, but I believe at that time when early timer is called, FDT blob might not be available yet.
- */
+unsigned long timer_early_get_rate(void);
#endif /* _TIMER_H_ */
[snip]
Regards, Bin
Hi Bin,
On 16 February 2016 at 02:21, Bin Meng bmeng.cn@gmail.com wrote:
Hi Simon,
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
In some cases the timer must be accessible before driver model is active. Examples include when using CONFIG_TRACE to trace U-Boot's execution before driver model is set up. Enable this option to use an early timer. These functions must be supported by your timer driver: timer_early_get_count() and timer_early_get_rate().
Signed-off-by: Simon Glass sjg@chromium.org
drivers/timer/Kconfig | 10 ++++++++++ include/timer.h | 21 +++++++++++++++++++++ lib/time.c | 28 +++++++++++++++++++++------- 3 files changed, 52 insertions(+), 7 deletions(-)
diff --git a/drivers/timer/Kconfig b/drivers/timer/Kconfig index ff65a73..cb18f12 100644 --- a/drivers/timer/Kconfig +++ b/drivers/timer/Kconfig @@ -9,6 +9,16 @@ config TIMER will be used. The timer is usually a 32 bits free-running up counter. There may be no real tick, and no timer interrupt.
+config TIMER_EARLY
bool "Allow timer to be used early in U-Boot"depends on TIMERhelpIn some cases the timer must be accessible before driver model isactive. Examples include when using CONFIG_TRACE to trace U-Boot'sexecution before driver model is set up. Enable this option touse an early timer. These functions must be supported by your timerdriver: timer_early_get_count() and timer_early_get_rate().config ALTERA_TIMER bool "Altera timer support" depends on TIMER diff --git a/include/timer.h b/include/timer.h index f14725c..a503bfd 100644 --- a/include/timer.h +++ b/include/timer.h @@ -67,4 +67,25 @@ struct timer_dev_priv { unsigned long clock_rate; };
+/**
- timer_early_get_count() - Implement timer_get_count() before driver model
- If CONFIG_TIMER_EARLY is enabled, this function wil be called to return
- the current timer value before the proper driver model timer is ready.
- It should be implemented by one of the timer values. This is mostly useful
- for tracing.
- */
+u64 timer_early_get_count(void);
+/**
- timer_early_get_rate() - Get the timer rate before driver model
- If CONFIG_TIMER_EARLY is enabled, this function wil be called to return
- the current timer rate in Hz before the proper driver model timer is ready.
nits: two spaces between 'rate' and 'in'
- It should be implemented by one of the timer values. This is mostly useful
- for tracing. This corresponds to the clock_rate value in struct
- timer_dev_priv.
Is this supposed to be a hard-codeded value returned by the timer driver? The timer-uclass driver gets this clock rate from device tree, but I believe at that time when early timer is called, FDT blob might not be available yet.
Possibly, although the FDT is available pretty early. I doubt anyone would need to call the timer that early.
- */
+unsigned long timer_early_get_rate(void);
#endif /* _TIMER_H_ */
[snip]
Regards, Bin
Regards, Simon
When using the early timer, we need to manually trigger setting up the real timer. This will not happen automatically. Do this immediately after starting driver model.
Signed-off-by: Simon Glass sjg@chromium.org ---
common/board_f.c | 6 ++++++ common/board_r.c | 14 ++++++++++++-- 2 files changed, 18 insertions(+), 2 deletions(-)
diff --git a/common/board_f.c b/common/board_f.c index a960144..622093a 100644 --- a/common/board_f.c +++ b/common/board_f.c @@ -45,6 +45,7 @@ #include <post.h> #include <spi.h> #include <status_led.h> +#include <timer.h> #include <trace.h> #include <video.h> #include <watchdog.h> @@ -805,6 +806,11 @@ static int initf_dm(void) if (ret) return ret; #endif +#ifdef CONFIG_TIMER_EARLY + ret = dm_timer_init(); + if (ret) + return ret; +#endif
return 0; } diff --git a/common/board_r.c b/common/board_r.c index 6c23865..52a9b26 100644 --- a/common/board_r.c +++ b/common/board_r.c @@ -46,6 +46,7 @@ #include <serial.h> #include <spi.h> #include <stdio_dev.h> +#include <timer.h> #include <trace.h> #include <watchdog.h> #ifdef CONFIG_CMD_AMBAPP @@ -312,13 +313,22 @@ static int initr_noncached(void) #ifdef CONFIG_DM static int initr_dm(void) { + int ret; + /* Save the pre-reloc driver model and start a new one */ gd->dm_root_f = gd->dm_root; gd->dm_root = NULL; -#ifdef CONFIG_TIMER + ret = dm_init_and_scan(false); + if (ret) + return ret; +#ifdef CONFIG_TIMER_EARLY gd->timer = NULL; + ret = dm_timer_init(); + if (ret) + return ret; #endif - return dm_init_and_scan(false); + + return 0; } #endif
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
When using the early timer, we need to manually trigger setting up the real timer. This will not happen automatically. Do this immediately after starting driver model.
Signed-off-by: Simon Glass sjg@chromium.org
common/board_f.c | 6 ++++++ common/board_r.c | 14 ++++++++++++-- 2 files changed, 18 insertions(+), 2 deletions(-)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
Add support for the early timer so we can use tracing with sandbox again.
Signed-off-by: Simon Glass sjg@chromium.org ---
drivers/timer/sandbox_timer.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/timer/sandbox_timer.c b/drivers/timer/sandbox_timer.c index a8da936..4537c82 100644 --- a/drivers/timer/sandbox_timer.c +++ b/drivers/timer/sandbox_timer.c @@ -18,9 +18,19 @@ void sandbox_timer_add_offset(unsigned long offset) sandbox_timer_offset += offset; }
-static int sandbox_timer_get_count(struct udevice *dev, u64 *count) +u64 notrace timer_early_get_count(void) { - *count = os_get_nsec() / 1000 + sandbox_timer_offset * 1000; + return os_get_nsec() / 1000 + sandbox_timer_offset * 1000; +} + +unsigned long notrace timer_early_get_rate(void) +{ + return 1000000; +} + +static notrace int sandbox_timer_get_count(struct udevice *dev, u64 *count) +{ + *count = timer_early_get_count();
return 0; }
Hi Simon,
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
Add support for the early timer so we can use tracing with sandbox again.
Signed-off-by: Simon Glass sjg@chromium.org
drivers/timer/sandbox_timer.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/timer/sandbox_timer.c b/drivers/timer/sandbox_timer.c index a8da936..4537c82 100644 --- a/drivers/timer/sandbox_timer.c +++ b/drivers/timer/sandbox_timer.c @@ -18,9 +18,19 @@ void sandbox_timer_add_offset(unsigned long offset) sandbox_timer_offset += offset; }
-static int sandbox_timer_get_count(struct udevice *dev, u64 *count) +u64 notrace timer_early_get_count(void) {
*count = os_get_nsec() / 1000 + sandbox_timer_offset * 1000;
return os_get_nsec() / 1000 + sandbox_timer_offset * 1000;+}
+unsigned long notrace timer_early_get_rate(void) +{
return 1000000;
Hard-coded?
+}
+static notrace int sandbox_timer_get_count(struct udevice *dev, u64 *count) +{
*count = timer_early_get_count(); return 0;}
Regards, Bin
Hi Bin
On 16 February 2016 at 02:22, Bin Meng bmeng.cn@gmail.com wrote:
Hi Simon,
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
Add support for the early timer so we can use tracing with sandbox again.
Signed-off-by: Simon Glass sjg@chromium.org
drivers/timer/sandbox_timer.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/timer/sandbox_timer.c b/drivers/timer/sandbox_timer.c index a8da936..4537c82 100644 --- a/drivers/timer/sandbox_timer.c +++ b/drivers/timer/sandbox_timer.c @@ -18,9 +18,19 @@ void sandbox_timer_add_offset(unsigned long offset) sandbox_timer_offset += offset; }
-static int sandbox_timer_get_count(struct udevice *dev, u64 *count) +u64 notrace timer_early_get_count(void) {
*count = os_get_nsec() / 1000 + sandbox_timer_offset * 1000;
return os_get_nsec() / 1000 + sandbox_timer_offset * 1000;+}
+unsigned long notrace timer_early_get_rate(void) +{
return 1000000;Hard-coded?
Yes - I'll convert it to a constant. But since the OS calls use nanoseconds and want to convert to milliseconds it is hard-coded.
+}
+static notrace int sandbox_timer_get_count(struct udevice *dev, u64 *count) +{
*count = timer_early_get_count(); return 0;}
Regards, Bin
Regards, Simon
This has got out of order: fix it.
Signed-off-by: Simon Glass sjg@chromium.org ---
configs/sandbox_defconfig | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/configs/sandbox_defconfig b/configs/sandbox_defconfig index b5b81ca..8b878e2 100644 --- a/configs/sandbox_defconfig +++ b/configs/sandbox_defconfig @@ -1,10 +1,13 @@ CONFIG_SYS_MALLOC_F_LEN=0x2000 CONFIG_PCI=y CONFIG_DEFAULT_DEVICE_TREE="sandbox" -CONFIG_DM_PCI_COMPAT=y CONFIG_FIT=y CONFIG_FIT_VERBOSE=y CONFIG_FIT_SIGNATURE=y +CONFIG_BOOTSTAGE=y +CONFIG_BOOTSTAGE_REPORT=y +CONFIG_CONSOLE_RECORD=y +CONFIG_CONSOLE_RECORD_OUT_SIZE=0x1000 # CONFIG_CMD_ELF is not set # CONFIG_CMD_IMLS is not set # CONFIG_CMD_FLASH is not set @@ -12,14 +15,10 @@ CONFIG_CMD_REMOTEPROC=y CONFIG_CMD_GPIO=y # CONFIG_CMD_SETEXPR is not set CONFIG_CMD_SOUND=y -CONFIG_BOOTSTAGE=y -CONFIG_BOOTSTAGE_REPORT=y CONFIG_CMD_PMIC=y CONFIG_CMD_REGULATOR=y CONFIG_CMD_TPM=y CONFIG_CMD_TPM_TEST=y -CONFIG_CONSOLE_RECORD=y -CONFIG_CONSOLE_RECORD_OUT_SIZE=0x1000 CONFIG_OF_CONTROL=y CONFIG_OF_HOSTFILE=y CONFIG_REGMAP=y @@ -52,6 +51,7 @@ CONFIG_SPI_FLASH_SST=y CONFIG_SPI_FLASH_WINBOND=y CONFIG_DM_ETH=y CONFIG_DM_PCI=y +CONFIG_DM_PCI_COMPAT=y CONFIG_PCI_SANDBOX=y CONFIG_PINCTRL=y CONFIG_PINCONF=y
Enable this so that tracing works with sandbox.
Signed-off-by: Simon Glass sjg@chromium.org ---
configs/sandbox_defconfig | 1 + 1 file changed, 1 insertion(+)
diff --git a/configs/sandbox_defconfig b/configs/sandbox_defconfig index 8b878e2..02534bf 100644 --- a/configs/sandbox_defconfig +++ b/configs/sandbox_defconfig @@ -68,6 +68,7 @@ CONFIG_SOUND=y CONFIG_SOUND_SANDBOX=y CONFIG_SANDBOX_SPI=y CONFIG_TIMER=y +CONFIG_TIMER_EARLY=y CONFIG_SANDBOX_TIMER=y CONFIG_TPM_TIS_SANDBOX=y CONFIG_USB=y
On Mon, Feb 15, 2016 at 9:36 AM, Simon Glass sjg@chromium.org wrote:
Enable this so that tracing works with sandbox.
Signed-off-by: Simon Glass sjg@chromium.org
configs/sandbox_defconfig | 1 + 1 file changed, 1 insertion(+)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
Add a little more debugging to help when things go wrong.
Signed-off-by: Simon Glass sjg@chromium.org ---
drivers/mtd/spi/sandbox.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
diff --git a/drivers/mtd/spi/sandbox.c b/drivers/mtd/spi/sandbox.c index 895604d..ec07be7 100644 --- a/drivers/mtd/spi/sandbox.c +++ b/drivers/mtd/spi/sandbox.c @@ -142,13 +142,15 @@ static int sandbox_sf_probe(struct udevice *dev) if (bus->seq < CONFIG_SANDBOX_SPI_MAX_BUS) spec = state->spi[bus->seq][cs].spec; if (!spec) { + debug("%s: No spec found for bus %d, cs %d\n", + __func__, bus->seq, cs); ret = -ENOENT; goto error; }
file = strchr(spec, ':'); if (!file) { - printf("sandbox_sf: unable to parse file\n"); + printf("%s: unable to parse file\n", __func__); ret = -EINVAL; goto error; } @@ -174,7 +176,7 @@ static int sandbox_sf_probe(struct udevice *dev) break; } if (!data->name) { - printf("sandbox_sf: unknown flash '%*s'\n", (int)idname_len, + printf("%s: unknown flash '%*s'\n", __func__, (int)idname_len, spec); ret = -EINVAL; goto error; @@ -186,7 +188,7 @@ static int sandbox_sf_probe(struct udevice *dev) sbsf->fd = os_open(pdata->filename, 02); if (sbsf->fd == -1) { free(sbsf); - printf("sandbox_sf: unable to open file '%s'\n", + printf("%s: unable to open file '%s'\n", __func__, pdata->filename); ret = -EIO; goto error; @@ -553,6 +555,9 @@ static int sandbox_cmdline_cb_spi_sf(struct sandbox_state *state, * yet. Perhaps we can figure something out. */ state->spi[bus][cs].spec = spec; + debug("%s: Setting up spec '%s' for bus %ld, cs %ld\n", __func__, + spec, bus, cs); + return 0; } SANDBOX_CMDLINE_OPT(spi_sf, 1, "connect a SPI flash: <bus>:<cs>:<id>:<file>"); @@ -671,6 +676,8 @@ int dm_scan_other(bool pre_reloc_only) __func__, busnum, cs); return ret; } + debug("%s: Setting up spec '%s' for bus %d, cs %d\n", + __func__, spec, busnum, cs); } } }
On 15 February 2016 at 07:06, Simon Glass sjg@chromium.org wrote:
Add a little more debugging to help when things go wrong.
Signed-off-by: Simon Glass sjg@chromium.org
Reviewed-by: Jagan Teki jteki@openedev.com Tested-by: Jagan Teki jteki@openedev.com
We must not free data that is managed by driver mode. Remove this line, which is a hangover from the pre-driver-model code.
This fixes a problem where 'sf probe' crashes U-Boot if the backing file for the SPI flash cannot be found.
Signed-off-by: Simon Glass sjg@chromium.org ---
drivers/mtd/spi/sandbox.c | 1 - 1 file changed, 1 deletion(-)
diff --git a/drivers/mtd/spi/sandbox.c b/drivers/mtd/spi/sandbox.c index ec07be7..53470b9 100644 --- a/drivers/mtd/spi/sandbox.c +++ b/drivers/mtd/spi/sandbox.c @@ -187,7 +187,6 @@ static int sandbox_sf_probe(struct udevice *dev)
sbsf->fd = os_open(pdata->filename, 02); if (sbsf->fd == -1) { - free(sbsf); printf("%s: unable to open file '%s'\n", __func__, pdata->filename); ret = -EIO;
On 15 February 2016 at 07:06, Simon Glass sjg@chromium.org wrote:
We must not free data that is managed by driver mode. Remove this line, which is a hangover from the pre-driver-model code.
This fixes a problem where 'sf probe' crashes U-Boot if the backing file for the SPI flash cannot be found.
Signed-off-by: Simon Glass sjg@chromium.org
Reviewed-by: Jagan Teki jteki@openedev.com Tested-by: Jagan Teki jteki@openedev.com
On Sun, Feb 14, 2016 at 06:36:58PM -0700, Simon Glass wrote:
We must not free data that is managed by driver mode. Remove this line, which is a hangover from the pre-driver-model code.
This fixes a problem where 'sf probe' crashes U-Boot if the backing file for the SPI flash cannot be found.
Signed-off-by: Simon Glass sjg@chromium.org
Reviewed-by: Tom Rini trini@konsulko.com
When an error number is provided we should use it, not change it. This fixes the SPI and SPI flash tests.
One of these is long-standing. The other seems to have been introduced by commit 1e90d9fd (sf: Move read_id code to sf_ops).
Signed-off-by: Simon Glass sjg@chromium.org Fixes: 1e90d9fd (sf: Move read_id code to sf_ops) ---
drivers/mtd/spi/sf_probe.c | 4 +--- drivers/mtd/spi/spi_flash.c | 2 +- 2 files changed, 2 insertions(+), 4 deletions(-)
diff --git a/drivers/mtd/spi/sf_probe.c b/drivers/mtd/spi/sf_probe.c index daa1d5b..7b29637 100644 --- a/drivers/mtd/spi/sf_probe.c +++ b/drivers/mtd/spi/sf_probe.c @@ -42,10 +42,8 @@ static int spi_flash_probe_slave(struct spi_flash *flash) }
ret = spi_flash_scan(flash); - if (ret) { - ret = -EINVAL; + if (ret) goto err_read_id; - }
#ifdef CONFIG_SPI_FLASH_MTD ret = spi_flash_mtd_register(flash); diff --git a/drivers/mtd/spi/spi_flash.c b/drivers/mtd/spi/spi_flash.c index 8a60c72..5c902f5 100644 --- a/drivers/mtd/spi/spi_flash.c +++ b/drivers/mtd/spi/spi_flash.c @@ -980,7 +980,7 @@ int spi_flash_scan(struct spi_flash *flash) ret = spi_flash_cmd(spi, CMD_READ_ID, idcode, sizeof(idcode)); if (ret) { printf("SF: Failed to get idcodes\n"); - return -EINVAL; + return ret; }
#ifdef DEBUG
On 15 February 2016 at 07:06, Simon Glass sjg@chromium.org wrote:
When an error number is provided we should use it, not change it. This fixes the SPI and SPI flash tests.
One of these is long-standing. The other seems to have been introduced by commit 1e90d9fd (sf: Move read_id code to sf_ops).
Signed-off-by: Simon Glass sjg@chromium.org Fixes: 1e90d9fd (sf: Move read_id code to sf_ops)
Reviewed-by: Jagan Teki jteki@openedev.com Tested-by: Jagan Teki jteki@openedev.com
These are working correctly again, so re-enable them.
Signed-off-by: Simon Glass sjg@chromium.org ---
test/dm/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/test/dm/Makefile b/test/dm/Makefile index d4f3f22..fd0198f 100644 --- a/test/dm/Makefile +++ b/test/dm/Makefile @@ -27,8 +27,8 @@ obj-y += regmap.o obj-$(CONFIG_REMOTEPROC) += remoteproc.o obj-$(CONFIG_RESET) += reset.o obj-$(CONFIG_DM_RTC) += rtc.o -#obj-$(CONFIG_DM_SPI_FLASH) += sf.o -#obj-$(CONFIG_DM_SPI) += spi.o +obj-$(CONFIG_DM_SPI_FLASH) += sf.o +obj-$(CONFIG_DM_SPI) += spi.o obj-y += syscon.o obj-$(CONFIG_DM_USB) += usb.o obj-$(CONFIG_DM_PMIC) += pmic.o
On 15 February 2016 at 07:07, Simon Glass sjg@chromium.org wrote:
These are working correctly again, so re-enable them.
Signed-off-by: Simon Glass sjg@chromium.org
Reviewed-by: Jagan Teki jteki@openedev.com Tested-by: Jagan Teki jteki@openedev.com
On 02/14/2016 06:36 PM, Simon Glass wrote:
Recent changes have broken the FIT and vboot tests. Also the SPI tests have been wrong since before the last release and were disabled.
This series collects together the required fixes.
Note: The FIT and vboot tests are hard to run (in that each requires manual effort). At some point we should be able to bring these into Stephen Warren's test framework.
The series, Tested-by: Stephen Warren swarren@nvidia.com (via ./test/py/test.py --bd sandbox --build)
participants (5)
-
Bin Meng -
Jagan Teki -
Simon Glass -
Stephen Warren -
Tom Rini