Calling the callback handler with spinlocks in the tasklet held leads
to deadlock when dmaengine functions are called:
BUG: spinlock lockup on CPU#0, sh/417, c1870a08
Backtrace:
...
[<c017b408>] (do_raw_spin_lock+0x0/0x154) from [<c02c4b98>] (_raw_spin_lock_irqsave+0x54/0x60)
[<c02c4b44>] (_raw_spin_lock_irqsave+0x0/0x60) from [<c01f5828>] (pl08x_prep_channel_resources+0x718/0x8b4)
[<c01f5110>] (pl08x_prep_channel_resources+0x0/0x8b4) from [<c01f5bb4>] (pl08x_prep_slave_sg+0x120/0x19c)
[<c01f5a94>] (pl08x_prep_slave_sg+0x0/0x19c) from [<c01be7a0>] (pl011_dma_tx_refill+0x164/0x224)
[<c01be63c>] (pl011_dma_tx_refill+0x0/0x224) from [<c01bf1c8>] (pl011_dma_tx_callback+0x7c/0xc4)
[<c01bf14c>] (pl011_dma_tx_callback+0x0/0xc4) from [<c01f4d34>] (pl08x_tasklet+0x60/0x368)
[<c01f4cd4>] (pl08x_tasklet+0x0/0x368) from [<c004d978>] (tasklet_action+0xa0/0x100)
Dan quoted the documentation:
> 2/ Completion callback routines cannot submit new operations. This
> results in recursion in the synchronous case and spin_locks being
> acquired twice in the asynchronous case.
but then followed up to say:
> I should clarify, this is the async_memcpy() api requirement which is
> not used outside of md/raid5. DMA drivers can and do allow new
> submissions from callbacks, and the ones that do so properly move the
> callback outside of the driver lock.
So let's fix it by moving the callback out of the spinlocked region.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>