Skip to contentSkip to conversation

review MPSC driver

DISCUSSION Randy.Dunlap 5 participants 5 messages Sep 16, 2004 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

| http://www.uwsg.iu.edu/hypermail/linux/kernel/0408.3/1549.html

Hi Mark,

1. Do you realize that a version of the driver is already in the -mm patchset?

2. + depends on PPC32 && MV64X60

Where is MV64X60 defined?

3. + select SERIAL_CORE
+ select SERIAL_CORE_CONSOLE

Please don't use "select". Use "depends on" instead.

4. + * Author: Mark A. Greer <mgreer@xxxxxxxxxx>

Put a real email address or remove it.

5. +#include <asm/mv64x60.h>

Where is this file? Does this driver build?

6. style:
+ if (pi->brg_can_tune) {
+ MPSC_MOD_FIELD_M(pi, brg, BRG_BCR, 1, 25, 0);
+ }
Has unneeded braces (in several places).

7. style:
+ return;
+}
Lots of void functions with "return" that is not needed.

8. Why use 'volatile' here? Have you read the Linus volatile rant?

+static inline void
+mpsc_sdma_set_tx_ring(struct mpsc_port_info *pi,
+ volatile struct mpsc_tx_desc *txre_p)
+{

9. put in mpsc.h:
+ static void mpsc_free_ring_mem(struct mpsc_port_info *pi);
+ static void mpsc_start_rx(struct mpsc_port_info *pi);

10. in the interrupt handler, if rx happened, tx intr not checked. Is that intentional?

+ if (mpsc_rx_intr(pi, regs) || mpsc_tx_intr(pi))
+ rc = IRQ_HANDLED;

11. In mpsc_verify_port(), if -EINVAL is ever set, the others are wasted checks.

12. What's the rationale for having both mpsc_console_init() and mpsc_late_console_init() ?

13. register_serial() and unregister_serial(): names are a bit too generic -- they sound like serial subsystem functions. and why are they exported? what else uses them?

14. mpsc.h: don't define MIN(), #include <linux/kernel.h> and use its min() macro.

15. Run it thru sparse for warnings.

--
~Randy

#2 lore raw
Quoted 10 lines · Randy.Dunlap

| http://www.uwsg.iu.edu/hypermail/linux/kernel/0408.3/1549.html

Hi Mark,
...
3. + select SERIAL_CORE
+ select SERIAL_CORE_CONSOLE

Please don't use "select". Use "depends on" instead.
...

That's a silly suggestion since none of these options are user visible.

Quoted 1 line

~Randy

cu
Adrian

Signature · 4 lines
       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed
#3 lore raw

On Thu, 16 Sep 2004 10:10:17 +0200 Adrian Bunk wrote:

| On Wed, Sep 15, 2004 at 09:43:01PM -0700, Randy.Dunlap wrote:
| > 
| > | http://www.uwsg.iu.edu/hypermail/linux/kernel/0408.3/1549.html
| > 
| > Hi Mark,
| >...
| > 3.  + select SERIAL_CORE
| >     + select SERIAL_CORE_CONSOLE
| > 
| > Please don't use "select".  Use "depends on" instead.
| >...
| 
| That's a silly suggestion since none of these options are user visible.

and Russell King wrote about the same text:

| This is actually (one of the few) correct uses of select.  These two
| symbols are *not* user visible, and are derived from the configuration
| settings of the hardware drivers.

Thanks for your comments. Any other comments on the driver?

--
~Randy

#4 lore raw
In reply to Randy.Dunlap
Quoted 4 lines · Randy.Dunlap

3. + select SERIAL_CORE
+ select SERIAL_CORE_CONSOLE

Please don't use "select". Use "depends on" instead.

This is actually (one of the few) correct uses of select. These two symbols are *not* user visible, and are derived from the configuration settings of the hardware drivers.

Signature · 4 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core
#5 lore raw
In reply to Randy.Dunlap
Quoted 4 lines · Randy.Dunlap

| http://www.uwsg.iu.edu/hypermail/linux/kernel/0408.3/1549.html

Hi Mark,

Hi Randy. Thanks for putting time into reviewing and commenting on this.

Quoted 3 lines

1. Do you realize that a version of the driver is already in the -mm patchset?

Ahh, no... Hmm, I guess someone did notice the patch I sent after all. :)

Quoted 4 lines

2. + depends on PPC32 && MV64X60

Where is MV64X60 defined?

There is a separate patch that provides "core" support for the gt/mv64x60 host bridges (which the MPSC is on). I haven't provided the patch yet b/c I keep finding little bugs in it. For modularity and to keep patch sizes smaller, I'm submitting separate patches for the core and for the mpsc driver. I had to draw the line somewhere between the driver and the core. It so happened that this file fell on the other side of that line. So, no, it won't compile until the core patch is submitted and accepted. However, if no one selects that driver--which they probably won't since the core support isn't there anyway--no harm done. I sent the mpsc driver patch before the core patch because it was more-or-less ready and so I could get the ball rolling WRT getting it accepted.

Quoted 6 lines

3. + select SERIAL_CORE
+ select SERIAL_CORE_CONSOLE

Please don't use "select". Use "depends on" instead.

Already discussed.

Quoted 4 lines

4. + * Author: Mark A. Greer <mgreer@xxxxxxxxxx>

Put a real email address or remove it.

Odd... It looks like the archive tried to do me a favor and "x-out" my domain so I don't get spam (I guess). The original patch has <[email protected]>, and that's reflected in the -mm patch and in other archives (e.g., http://lkml.org/lkml/2004/8/27/305). I should have pointed you to the lkml.org link instead.

Quoted 4 lines

5. +#include <asm/mv64x60.h>

Where is this file? Does this driver build?

See comments on 2.

Quoted 6 lines

6. style:

(quoted diff) +3 -0
+	if (pi->brg_can_tune) {
+		MPSC_MOD_FIELD_M(pi, brg, BRG_BCR, 1, 25, 0);
+	}

Has unneeded braces (in several places).

I will fix.

Quoted 5 lines

7. style:

(quoted diff) +2 -0
+	return;
+}

Lots of void functions with "return" that is not needed.

Well, I didn't notice anything in Documentation/CodingStyle on this so I assumed it was up to me. If there is a policy on this, I'll remove them. If its up to me, I'd prefer that they be left in--redundant, yes, but somehow I like the clarity.

Review

Quoted 7 lines

8. Why use 'volatile' here? Have you read the Linus volatile rant?

(quoted diff) +4 -0
+static inline void
+mpsc_sdma_set_tx_ring(struct mpsc_port_info *pi,
+		      volatile struct mpsc_tx_desc *txre_p)
+{

I have heard about it but I've not actually read it. I will read it and change whatever needs to be.

Review

Quoted 5 lines

9. put in mpsc.h:

(quoted diff) +2 -0
+	static void mpsc_free_ring_mem(struct mpsc_port_info *pi);
+	static void mpsc_start_rx(struct mpsc_port_info *pi);

I didn't put them in mpsc.h because they're never referenced outside of mpsc.c. I will move them unless someone else objects.

Review

Quoted 7 lines

10. in the interrupt handler, if rx happened, tx intr not checked. Is that intentional?

(quoted diff) +2 -0
+	if (mpsc_rx_intr(pi, regs) || mpsc_tx_intr(pi))
+		rc = IRQ_HANDLED;

Good point. I should have caught that one.

Quoted 3 lines

11. In mpsc_verify_port(), if -EINVAL is ever set, the others are wasted checks.

Another good point.

Quoted 4 lines

12. What's the rationale for having both mpsc_console_init() and mpsc_late_console_init() ?

I modeled that after the 8250 driver. I am not intimately familiar with all of the char/tty/serial infrastructure so I erred on the safe side and did what was done in the 8250 driver. I will look deeper to see if I can find a reason.

Quoted 4 lines

13. register_serial() and unregister_serial(): names are a bit too generic -- they sound like serial subsystem functions. and why are they exported? what else uses them?

AFAICT, they are standard interfaces for "add-on" files to call into "core" files like 8250_pci.c calls 8250.c's register_serial(). Since the mpsc doesn't really have any add-on files, I believe that I can get rid of them.

Quoted 3 lines

14. mpsc.h: don't define MIN(), #include <linux/kernel.h> and use its min() macro.

Yep, already pointed out to me by Russell. Will be fixed.

Quoted 3 lines

15. Run it thru sparse for warnings.

Will do.

BTW, I'm waiting on a minor number (for major 204) from lanana so I won't post a new patch until I get that (or I get no response in which case, I guess I'm forced to pick one).

Thanks again,

Mark