Message ID | 1702276972-41296-1-git-send-email-quic_qianyu@quicinc.com |
---|---|
Headers | show |
Series | bus: mhi: host: Add lock to avoid race when ringing channel DB | expand |
On 12/10/2023 11:42 PM, Qiang Yu wrote: > Ensure read and write locks for the channel are not taken in succession by > dropping the read lock from parse_xfer_event() such that a callback given > to client can potentially queue buffers and acquire the write lock in that > process. Any queueing of buffers should be done without channel read lock > acquired as it can result in multiple locks and a soft lockup. > > Signed-off-by: Qiang Yu <quic_qianyu@quicinc.com> Seems to work fine for AIC100 Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Tested-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
On Mon, Dec 11, 2023 at 02:42:52PM +0800, Qiang Yu wrote: > Ensure read and write locks for the channel are not taken in succession by > dropping the read lock from parse_xfer_event() such that a callback given > to client can potentially queue buffers and acquire the write lock in that > process. Any queueing of buffers should be done without channel read lock > acquired as it can result in multiple locks and a soft lockup. > Cc: <stable@vger.kernel.org> # 5.7 Fixes: 1d3173a3bae7 ("bus: mhi: core: Add support for processing events from client device") > Signed-off-by: Qiang Yu <quic_qianyu@quicinc.com> Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> - Mani > --- > drivers/bus/mhi/host/main.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c > index 32021fe..25f98d6 100644 > --- a/drivers/bus/mhi/host/main.c > +++ b/drivers/bus/mhi/host/main.c > @@ -642,6 +642,8 @@ static int parse_xfer_event(struct mhi_controller *mhi_cntrl, > mhi_del_ring_element(mhi_cntrl, tre_ring); > local_rp = tre_ring->rp; > > + read_unlock_bh(&mhi_chan->lock); > + > /* notify client */ > mhi_chan->xfer_cb(mhi_chan->mhi_dev, &result); > > @@ -667,6 +669,8 @@ static int parse_xfer_event(struct mhi_controller *mhi_cntrl, > kfree(buf_info->cb_buf); > } > } > + > + read_lock_bh(&mhi_chan->lock); > } > break; > } /* CC_EOT */ > -- > 2.7.4 > >
On Mon, Dec 11, 2023 at 02:42:50PM +0800, Qiang Yu wrote: > > 1. We need a write lock in mhi_gen_tre otherwise there is race of the WP > used for ringing channel DB between mhi_queue and M0 transition. > 2. We can not invoke local_bh_enable() when irqs are disabled, so move > read_lock_irqsave() under the mhi_gen_tre() since we add write_lock_bh() in > mhi_gen_tre(). > 3. Unlock xfer_cb to prevent potential lockup > Applied to mhi-next! - Mani > v1 -> v2: > Added write_unlock_bh(&mhi_chan->lock) in mhi_gen_tre() before return > because of error process. > > v2 -> v3: > 1. split protecting WP and unlocking xfer_cb into two patches > 2. Add a new patch to stop processing buffer and eventof a disabled or > stopped channel. > > v3 -> v4: > 1. Modify commit message > 2. Add unlock operation before return error > > v4 -> v5: > 1. Squash "protecting WP" and "Take irqsave lock" into one patch > 2. Drop patch 3/4 of patch v4 > > Bhaumik Bhatt (1): > bus: mhi: host: Add spinlock to protect WP access when queueing TREs > > Qiang Yu (1): > bus: mhi: host: Drop chan lock before queuing buffers > > drivers/bus/mhi/host/main.c | 26 +++++++++++++++++--------- > 1 file changed, 17 insertions(+), 9 deletions(-) > > -- > 2.7.4 > >