n_gsm.c: add tx_lock in gsm_send

From: xiaojin <[email protected]>
Date: Wed, 19 Dec 2012 11:53:43 +0800
Subject: [PATCH] n_gsm.c: add tx_lock in gsm_send

All the call to gsm->output should be in the tx_lock, that could avoid potential race from MUX level. But we have no tx_lock in gsm_send.

This patch is to add tx_lock in gsm_send.

---
 drivers/tty/n_gsm.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

index dcc0430..4572117 100644
@@ -573,6 +573,7 @@ static void gsm_send(struct gsm_mux *gsm, int addr,

int cr, int control)
int len;
u8 cbuf[10];
u8 ibuf[3];
+ unsigned long flags;

switch (gsm->encoding) {
case 0:
@@ -602,7 +603,9 @@ static void gsm_send(struct gsm_mux *gsm, int addr, int cr, int control)
WARN_ON(1);
return;
}
+ spin_lock_irqsave(&gsm->tx_lock, flags);
gsm->output(gsm, cbuf, len);
+ spin_unlock_irqrestore(&gsm->tx_lock, flags);
gsm_print_packet("-->", addr, cr, control, NULL, 0);
}

1.7.1

gsm->output calls the transmit method of the underlying tty driver. We can't do that with interrupts off as some drivers expect to be able to sleep in their output paths.

Alan

Alan,

Thanks. But the comment makes me confused. As we see, gsm->output is called by gsm_data_kick too, and it's in the tx_lock...

Best regards,
Jin Xiao

gsm->output calls the transmit method of the underlying tty driver. We
can't do that with interrupts off as some drivers expect to be able to sleep in their output paths.

Alan

That would be a bug too or I guess we could finally give in on trying to keep tty write paths not using spin_lock_irq and fix any oddments that blow up. That might actually be the right thing to do as its caused problems in other places too, and the USB tty drivers are now well behaved.

Your patch doesn't make it any worse however so I agree it might as well go in.

Alan