Skip to contentSkip to conversation

cyclades: Prepare for relaxed locking in callers

PATCH In review Alan Cox 2 participants 4 messages Feb 20, 2008 Char and misc drivers

lore
Share & tools
lore

Alan Cox

author
#0 lore raw

PATCH 1/1 cyclades: Prepare for relaxed locking in callers

Basically wrap it in lock_kernel where it is hard to prove the locking is ok.

drivers/char/cyclades.c +17 -12
diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.25-rc2-mm1/drivers/char/cyclades.c linux-2.6.25-rc2-mm1/drivers/char/cyclades.c
@@ -3464,6 +3464,8 @@
 	if (serial_paranoia_check(info, tty->name, __FUNCTION__))
 		return -ENODEV;
 
+	lock_kernel();
+	
 	card = info->card;
 	channel = info->line - card->first_line;
 	if (!IS_CYC_Z(*card)) {
@@ -3506,10 +3508,12 @@
 				((lstatus & C_RS_CTS) ? TIOCM_CTS : 0);
 		} else {
 			result = 0;
+			unlock_kernel();
 			return -ENODEV;
 		}
 
 	}
+	unlock_kernel();
 	return result;
 }				/* cy_tiomget */
 
@@ -3880,6 +3884,7 @@
 	printk(KERN_DEBUG "cyc:cy_ioctl ttyC%d, cmd = %x arg = %lx\n",
 		info->line, cmd, arg);
 #endif
+	lock_kernel();
 
 	switch (cmd) {
 	case CYGETMON:
@@ -3988,47 +3993,47 @@
 		p_cuser = argp;
 		ret_val = put_user(cnow.cts, &p_cuser->cts);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.dsr, &p_cuser->dsr);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.rng, &p_cuser->rng);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.dcd, &p_cuser->dcd);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.rx, &p_cuser->rx);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.tx, &p_cuser->tx);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.frame, &p_cuser->frame);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.overrun, &p_cuser->overrun);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.parity, &p_cuser->parity);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.brk, &p_cuser->brk);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = put_user(cnow.buf_overrun, &p_cuser->buf_overrun);
 		if (ret_val)
-			return ret_val;
+			break;
 		ret_val = 0;
 		break;
 	default:
 		ret_val = -ENOIOCTLCMD;
 	}
+	unlock_kernel();
 
 #ifdef CY_DEBUG_OTHER
 	printk(KERN_DEBUG "cyc:cy_ioctl done\n");
 #endif
-
 	return ret_val;
 }				/* cy_ioctl */
 
#1 lore raw

Alan> Basically wrap it in lock_kernel where it is hard to prove the Alan> locking is ok.

I've got cyclades cards, both ISA and Serial. Do you want/need any specific tests? Or should I just send you (or your deputy) the ISA card for your collection?

John

Alan> Signed-off-by: Alan Cox <[email protected]>

Alan> diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.25-rc2-mm1/drivers/char/cyclades.c linux-2.6.25-rc2-mm1/drivers/char/cyclades.c
Alan> --- linux.vanilla-2.6.25-rc2-mm1/drivers/char/cyclades.c 2008-02-19 11:01:43.000000000 +0000
Alan> +++ linux-2.6.25-rc2-mm1/drivers/char/cyclades.c 2008-02-20 11:45:28.000000000 +0000
Alan> @@ -3464,6 +3464,8 @@
Alan> if (serial_paranoia_check(info, tty->name, __FUNCTION__))
Alan> return -ENODEV;

Alan> + lock_kernel();
Alan> +

Alan>  	card = info->card;
Alan>  	channel = info->line - card->first_line;
Alan>  	if (!IS_CYC_Z(*card)) {
Alan> @@ -3506,10 +3508,12 @@
Alan>  				((lstatus & C_RS_CTS) ? TIOCM_CTS : 0);
Alan>  		} else {
Alan>  			result = 0;

Alan> + unlock_kernel();
Alan> return -ENODEV;
Alan> }

Alan> }
Alan> + unlock_kernel();
Alan> return result;
Alan> } /* cy_tiomget */

Alan> @@ -3880,6 +3884,7 @@
Alan> printk(KERN_DEBUG "cyc:cy_ioctl ttyC%d, cmd = %x arg = %lx\n",
info-> line, cmd, arg);
Alan> #endif
Alan> + lock_kernel();

Alan> switch (cmd) {
Alan> case CYGETMON:
Alan> @@ -3988,47 +3993,47 @@

Alan>  		p_cuser = argp;
Alan>  		ret_val = put_user(cnow.cts, &p_cuser->cts);
Alan>  		if (ret_val)

Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.dsr, &p_cuser->dsr);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.rng, &p_cuser->rng);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.dcd, &p_cuser->dcd);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.rx, &p_cuser->rx);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.tx, &p_cuser->tx);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.frame, &p_cuser->frame);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.overrun, &p_cuser->overrun);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.parity, &p_cuser->parity);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.brk, &p_cuser->brk);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;
Alan> ret_val = put_user(cnow.buf_overrun, &p_cuser->buf_overrun);
Alan> if (ret_val)
Alan> - return ret_val;
Alan> + break;

Alan>  		ret_val = 0;
Alan>  		break;
Alan>  	default:
Alan>  		ret_val = -ENOIOCTLCMD;
Alan>  	}

Alan> + unlock_kernel();

Alan>  #ifdef CY_DEBUG_OTHER
Alan>  	printk(KERN_DEBUG "cyc:cy_ioctl done\n");
Alan>  #endif

Alan> -
Alan> return ret_val;
Alan> } /* cy_ioctl */

Alan> --
Alan> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in Alan> the body of a message to [email protected]
Alan> More majordomo info at http://vger.kernel.org/majordomo-info.html
Alan> Please read the FAQ at http://www.tux.org/lkml/

Alan> !DSPAM:47bc8c2521682037060298!

author
#2 lore raw
Quoted 7 lines · On Wed, 20 Feb 2008 16:22:29 -0500 "John Stoffel" <[email protected]> wrote:

Alan> Basically wrap it in lock_kernel where it is hard to prove the Alan> locking is ok.

I've got cyclades cards, both ISA and Serial. Do you want/need any specific tests? Or should I just send you (or your deputy) the ISA card for your collection?

At the moment its fairly tricky to test unless you want to hack your tree up further. The patch prepares for stuff that I've not yet sent Andrew. I want to make sure we haven't broken anything before sending the small patches to fire it all up.

#3 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Alan" == Alan Cox <[email protected]> writes:

Alan> On Wed, 20 Feb 2008 16:22:29 -0500

Quoted 1 line · Alan> "John Stoffel" <[email protected]> wrote:
Quoted 1 line

Alan> Basically wrap it in lock_kernel where it is hard to prove the Alan> locking is ok.

Quoted 4 lines
Quoted 4 lines

I've got cyclades cards, both ISA and Serial. Do you want/need any specific tests? Or should I just send you (or your deputy) the ISA card for your collection?

Alan> At the moment its fairly tricky to test unless you want to hack Alan> your tree up further. The patch prepares for stuff that I've not Alan> yet sent Andrew. I want to make sure we haven't broken anything Alan> before sending the small patches to fire it all up.

Ok, please feel free to keep me in mind for testing. I've still got ISA box(en) here up and running and being used. As well as my Cyclom-Y PCI card in actual use still.

John