> On Nov 27, 2018, at 8:52 AM, Sean Christopherson <sean.j.christopherson@xxxxxxxxx> wrote:
>
> On Tue, Nov 27, 2018 at 09:36:03AM +0100, Heiko Carstens wrote:
>> Use pr_alert_once() instead of pr_alert() if page table misaccounting
>> has been detected.
>>
>> If this happens once it is very likely that there will be numerous
>> other occurrence as well, which would flood dmesg and the console with
>> hardly any added information. Therefore print the warning only once.
>>
>> Cc: Kirill A. Shutemov <kirill@xxxxxxxxxxxxx>
>> Cc: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
>> Signed-off-by: Heiko Carstens <heiko.carstens@xxxxxxxxxx>
>> ---
>> kernel/fork.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/fork.c b/kernel/fork.c
>> index 07cddff89c7b..c887e9eba89f 100644
>> --- a/kernel/fork.c
>> +++ b/kernel/fork.c
>> @@ -647,8 +647,8 @@ static void check_mm(struct mm_struct *mm)
>> }
>>
>> if (mm_pgtables_bytes(mm))
>> - pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
>> - mm_pgtables_bytes(mm));
>> + pr_alert_once("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
>> + mm_pgtables_bytes(mm));
>
> I found the print-always behavior to be useful when developing a driver
> that mucked with PTEs directly via vmf_insert_pfn() and had issues with
> racing against exit_mmap(). It was nice to be able to recompile only
> the driver and rely on dmesg to let me know when I messed up yet again.
>
> Would pr_alert_ratelimited() suffice?
Actually, I really like that idea.
There are certainly times when it is useful to see a cascade of messages, within reason;
one there are so many they overflow the dmesg buffer they're of limited usefulness.
Something like a pr_alert() that could rate limit to a preset value, perhaps a default of
fifty or so, could prove quite useful indeed without being an all or once choice.
William Kucharski
