[U-Boot] [RFC PATCH v3 0/2] env: Make environment loading log more clear
This is another attempt to make boot log better, without hackery this time. Basically here we just remove unwanted error messages, relying on the message from most deep API to be printed (like mmc subsystem). This approach *might* have its shortcomings, though: 1. With no "Failed" message, at some point we *can* end up with no error messages printed at all 2. Removing some collateral error messages *may* lead to loss of useful debug info in other use-cases (env_load() is not only user of those APIs).
That being said, at the moment this looks like most clean solution to cluttered log problem, as any other solution either will be hackish or will require some big architectural changes. If one of mentioned shortcomings occur, we can fix it ad hoc.
With this patch set applied we will see something like this:
Loading Environment from FAT... MMC: no card present Loading Environment from MMC... OK
instead of:
Loading Environment from FAT... MMC: no card present ** Bad device mmc 0 ** Failed (-5) Loading Environment from MMC... OK
Sam Protsenko (2): env: Don't print "Failed" error message disk: part: Remove redundant error message
disk/part.c | 1 - env/env.c | 4 +--- 2 files changed, 1 insertion(+), 4 deletions(-)
This only clutters the log with unnecessary details, as we already have all needed warnings by the time. Example:
Loading Environment from FAT... MMC: no card present ** Bad device mmc 0 ** Failed (-5)
Remove this "Failed" message to keep log short and clear.
Signed-off-by: Sam Protsenko semen.protsenko@linaro.org --- env/env.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/env/env.c b/env/env.c index 5c0842ac07..3ab4ec4237 100644 --- a/env/env.c +++ b/env/env.c @@ -196,9 +196,7 @@ int env_load(void)
printf("Loading Environment from %s... ", drv->name); ret = drv->load(); - if (ret) - printf("Failed (%d)\n", ret); - else + if (!ret) printf("OK\n");
if (!ret)
Underlying API should already print some meaningful error message, so this one is just brings more noise. E.g. we can see log like this:
MMC: no card present ** Bad device mmc 0 **
Obviously, second error message is unwanted. Let's remove it to make log more short and clear.
Signed-off-by: Sam Protsenko semen.protsenko@linaro.org --- disk/part.c | 1 - 1 file changed, 1 deletion(-)
diff --git a/disk/part.c b/disk/part.c index 9266a09ec3..0762a0750f 100644 --- a/disk/part.c +++ b/disk/part.c @@ -400,7 +400,6 @@ int blk_get_device_by_str(const char *ifname, const char *dev_hwpart_str,
*dev_desc = get_dev_hwpart(ifname, dev, hwpart); if (!(*dev_desc) || ((*dev_desc)->type == DEV_TYPE_UNKNOWN)) { - printf("** Bad device %s %s **\n", ifname, dev_hwpart_str); dev = -ENOENT; goto cleanup; }
On Fri, Jul 20, 2018 at 01:28:43AM +0300, Sam Protsenko wrote:
Underlying API should already print some meaningful error message, so this one is just brings more noise. E.g. we can see log like this:
MMC: no card present ** Bad device mmc 0 **Obviously, second error message is unwanted. Let's remove it to make log more short and clear.
Signed-off-by: Sam Protsenko semen.protsenko@linaro.org
disk/part.c | 1 - 1 file changed, 1 deletion(-)
diff --git a/disk/part.c b/disk/part.c index 9266a09ec3..0762a0750f 100644 --- a/disk/part.c +++ b/disk/part.c @@ -400,7 +400,6 @@ int blk_get_device_by_str(const char *ifname, const char *dev_hwpart_str,
*dev_desc = get_dev_hwpart(ifname, dev, hwpart); if (!(*dev_desc) || ((*dev_desc)->type == DEV_TYPE_UNKNOWN)) {
printf("** Bad device %s %s **\n", ifname, dev_hwpart_str);
We should move this to debug() I think.
On Fri, Jul 20, 2018 at 1:53 AM, Tom Rini trini@konsulko.com wrote:
On Fri, Jul 20, 2018 at 01:28:43AM +0300, Sam Protsenko wrote:
Underlying API should already print some meaningful error message, so this one is just brings more noise. E.g. we can see log like this:
MMC: no card present ** Bad device mmc 0 **Obviously, second error message is unwanted. Let's remove it to make log more short and clear.
Signed-off-by: Sam Protsenko semen.protsenko@linaro.org
disk/part.c | 1 - 1 file changed, 1 deletion(-)
diff --git a/disk/part.c b/disk/part.c index 9266a09ec3..0762a0750f 100644 --- a/disk/part.c +++ b/disk/part.c @@ -400,7 +400,6 @@ int blk_get_device_by_str(const char *ifname, const char *dev_hwpart_str,
*dev_desc = get_dev_hwpart(ifname, dev, hwpart); if (!(*dev_desc) || ((*dev_desc)->type == DEV_TYPE_UNKNOWN)) {
printf("** Bad device %s %s **\n", ifname, dev_hwpart_str); dev = -ENOENT; goto cleanup; }We should move this to debug() I think.
Yes, better to use debug() than remove it completely, thanks. So, if there is no further objections, I'm gonna set real patch series soon.
-- Tom
Dear Sam,
In message 20180719222843.28316-1-semen.protsenko@linaro.org you wrote:
- With no "Failed" message, at some point we *can* end up with no error messages printed at all
That would mean that we did not check the whole call tree as needed. It is not that complicated, or is it?
- Removing some collateral error messages *may* lead to loss of useful debug info in other use-cases (env_load() is not only user of those APIs).
I dislike the "can" and "may" parts here. If this is done thoroughly, we should know exactly that no such damage gets done.
As I mentioend before: if we get here, somewhere an error must have occurred. And in the error handling an error message (one) must be printed.
Best regards,
Wolfgang Denk
On Fri, Jul 20, 2018 at 4:27 PM, Wolfgang Denk wd@denx.de wrote:
Dear Sam,
In message 20180719222843.28316-1-semen.protsenko@linaro.org you wrote:
- With no "Failed" message, at some point we *can* end up with no error messages printed at all
That would mean that we did not check the whole call tree as needed. It is not that complicated, or is it?
I've checked the error chain for FAT/MMC env load: [1]. As for the rest of possible sources, frankly I didn't check, as I don't have such boards at my disposal anyway, and I may not be aware of all sources we have. That's why I've only corrected the path for loading from FAT, which I can test and can argue it's working right. I will add the comment in env_load() saying that error message must be printed from underlying subsystem, and it should be exactly one error. If we catch any related issues further, we can check it on the sport. I hope you're ok with this approach, as fixing all possible error cases for all possible sources in at once seems to be too complex task. And some further messages can appear further, if someone would manage to get such a patch merged. So let's start with clear requirement in env_load().
[1] https://pastebin.com/q6kgpprb
- Removing some collateral error messages *may* lead to loss of useful debug info in other use-cases (env_load() is not only user of those APIs).
I dislike the "can" and "may" parts here. If this is done thoroughly, we should know exactly that no such damage gets done.
Agree. Also, I like Tom's suggestion to move those to debug() instead of removing.
As I mentioend before: if we get here, somewhere an error must have occurred. And in the error handling an error message (one) must be printed.
Best regards,
Wolfgang Denk
-- DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de Software entities are more complex for their size than perhaps any other human construct because no two parts are alike. If they are, we make the two similar parts into a subroutine -- open or closed. In this respect, software systems differ profoundly from computers, buildings, or automobiles, where repeated elements abound. - Fred Brooks, Jr.
participants (3)
-
Sam Protsenko -
Tom Rini -
Wolfgang Denk