Re: [PATCH 4/4] drivers/xen/xenbus/xenbus_client.c: Improve the failure processing for __xenbus_switch_state()

[Chen Gang <gang.chen.5i5j@gmail.com>]

On 9/16/14 9:07, Boris Ostrovsky wrote:
> 
> On 09/14/2014 06:52 AM, Chen Gang wrote:
>> When failure occurs, need return failure code instead of 0, or the upper
>> caller will misunderstand.
>>
>> Also when retry for EAGAIN reason, better to schedule out for a while,
>> so can let others have chance to continue their tasks (especially,
>> their tasks are related EAGAIN under UP kernel).
>>
>> Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
>> ---
>>   drivers/xen/xenbus/xenbus_client.c | 9 ++++++---
>>   1 file changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/xen/xenbus/xenbus_client.c b/drivers/xen/xenbus/xenbus_client.c
>> index 620ffd2..fc8699b 100644
>> --- a/drivers/xen/xenbus/xenbus_client.c
>> +++ b/drivers/xen/xenbus/xenbus_client.c
>> @@ -36,6 +36,7 @@
>>   #include <linux/spinlock.h>
>>   #include <linux/vmalloc.h>
>>   #include <linux/export.h>
>> +#include <linux/delay.h>
>>   #include <asm/xen/hypervisor.h>
>>   #include <asm/xen/page.h>
>>   #include <xen/interface/xen.h>
>> @@ -196,7 +197,7 @@ again:
>>       err = xenbus_transaction_start(&xbt);
>>       if (err) {
>>           xenbus_switch_fatal(dev, depth, err, "starting transaction");
>> -        return 0;
>> +        return err;
>>       }
>>         err = xenbus_scanf(xbt, dev->nodename, "state", "%d", &current_state);
>> @@ -213,13 +214,15 @@ again:
>>   abort:
>>       err = xenbus_transaction_end(xbt, abort);
>>       if (err) {
>> -        if (err == -EAGAIN && !abort)
>> +        if (err == -EAGAIN && !abort) {
>> +            msleep(20);
> 
> I'd suggest you use schedule() here.
> 

For me, it is really better than msleep(20).

But just like the related maintainer said, we do not need it, since it
has already "have a rest" inside xenbus_transaction_end().


Thanks.
-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed