2009-01-11 10:02:58 spin_lock_irqsave in interrupt handler

Servaes Joordens (NETHERLANDS)

Message: 67765

Is it correct to have an spin_lock_irqsave in a interrup handler?

It is used in the ppi driver.

In my opinion this can lead to an problem:

Suppose an ioctl call is made and during this call the ppi dma interrupt occurs: then the interrupt blocks because the spin_lock is already locked by the ioctl call.

regards,

Servaes

QuoteReplyEditDelete

2009-01-11 10:09:04 Re: spin_lock_irqsave in interrupt handler

Mike Frysinger (UNITED STATES)

Message: 67766

it is invalid to sleep while holding a spinlock, so the issue you describe can never occur in a correctly written driver. i dont see any code in the ppi ioctl() function that will sleep (ignoring pr_debug() statements).

QuoteReplyEditDelete

2009-01-11 11:39:58 Re: spin_lock_irqsave in interrupt handler

Servaes Joordens (NETHERLANDS)

Message: 67767

Is the spinlock behaving as a mutex? In that case it will wait until the spin_lock is released in the ioctl function.

QuoteReplyEditDelete

2009-01-11 12:00:37 Re: spin_lock_irqsave in interrupt handler

Mike Frysinger (UNITED STATES)

Message: 67768

a spinlock is a spinlock, a mutex is a mutex. there is no sleeping by definition.

http://en.wikipedia.org/wiki/Spinlock

http://en.wikipedia.org/wiki/Mutual_exclusion