Hi Neil,
Thank you for a kernel patch and a makedumpfile patch.
Neil Horman wrote:
>> That aside we aren't currently exporting log_buf_len, so I don't
>> think this code works actually works.
>>
>> Neil can you add a comment in kernel/printk.c of the algorithm
>> necessary for external tools to decode the ring buffer?
>>
>> We need the comment because people working on kernel/printk.c
>> need to know what is happening and without having to review lots
>> of user space code, and we need the comment to verify that we
>> are exporting the right things.
>>
>
> Ok, as per Erics comment, I've written this. It applies on top of whats already
> in your tree Andrew. It adds some comments on the function in question so that
> anyone working on printk.c will know why we're exporting their symbols. It also
> modifies slightly the symbols we are exporting so that we can handle dmesg
> buffers that are longer than the standard PAGE_SIZE configuration, and lets us
> detect and handle buffer wraps.
>
> Thanks!
>
> printk.c | 12 +++++++++++-
> 1 file changed, 11 insertions(+), 1 deletion(-)
>
> Signed-off-by: Neil Horman <nhorman at tuxdriver.com>
>
> diff --git a/kernel/printk.c b/kernel/printk.c
> index dd5cc7b..6a1163b 100644
> --- a/kernel/printk.c
> +++ b/kernel/printk.c
> @@ -138,10 +138,20 @@ static int log_buf_len = __LOG_BUF_LEN;
> static unsigned logged_chars; /* Number of chars produced since last read+clear operation */
>
> #ifdef CONFIG_KEXEC
> +/*
> + * This appends the listed symbols to /proc/vmcoreinfo
> + * /proc/vmcoreinfo is used by various utiilties, like
> + * crash and makedumpfile to obtain access to symbols that
> + * are otherwise very difficult to locate. These symbols
> + * are specifically used so that utilities
> + * can access and extract the dmesg log from a vmcore file
> + * after a crash
> + */
> void log_buf_kexec_setup(void)
> {
> VMCOREINFO_SYMBOL(log_buf);
> - VMCOREINFO_SYMBOL(log_end);
> + VMCOREINFO_SYMBOL(log_buf_len);
> + VMCOREINFO_SYMBOL(logged_chars)
> }
> #endif
Please do not remove VMCOREINFO_SYMBOL(log_end), because makedumpfile
needs it for knowing the latest log offset in ring buffer of dmesg.
I updated a makedumpfile patch for handling ring buffer, and it is
possible to download the latest patch from the following site:
https://sourceforge.net/tracker2/download.php?group_id=178938&atid=887142&file_id=312113&aid=2521075
Thanks
Ken'ichi Ohmichi
