Skip to contentSkip to conversation

greybus: Use fallthrough pseudo-keyword

PATCH In review Gustavo A. R. Silva 6 participants 7 messages Jul 27, 2020 – Aug 5, 2020 Greybus

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 greybus: Use fallthrough pseudo-keyword

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

---
 drivers/greybus/es2.c       | 2 +-
 drivers/greybus/interface.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
drivers/greybus/es2.c +1 -1
index 366716f11b1a..1df6ab5d339d 100644
@@ -759,7 +759,7 @@ static int check_urb_status(struct urb *urb)
 	case -EOVERFLOW:
 		dev_err(dev, "%s: overflow actual length is %d\n",
 			__func__, urb->actual_length);
-		/* fall through */
+		fallthrough;
 	case -ECONNRESET:
 	case -ENOENT:
 	case -ESHUTDOWN:
drivers/greybus/interface.c +1 -1
index 67dbe6fda9a1..58ea374d8aaa 100644
@@ -1233,7 +1233,7 @@ int gb_interface_add(struct gb_interface *intf)
 	case GB_INTERFACE_TYPE_GREYBUS:
 		dev_info(&intf->dev, "GMP VID=0x%08x, PID=0x%08x\n",
 			 intf->vendor_id, intf->product_id);
-		/* fall-through */
+		fallthrough;
 	case GB_INTERFACE_TYPE_UNIPRO:
 		dev_info(&intf->dev, "DDBL1 Manufacturer=0x%08x, Product=0x%08x\n",
 			 intf->ddbl1_manufacturer_id,
Signature · 1 line
2.27.0
#1 lore raw
Quoted 4 lines · Gustavo A. R. Silva

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Thanks for the patch. It looks good, but it raises another question I'd like discussion on.

It seems that Johan likes default (or final) cases in switch statements without a "break" statement. Viresh and Bryan appear to be fond of this too.

It's pedantic, but I don't like that. Am I wrong?
--> Johan/Viresh/Bryan would you please comment?

If they aren't convincing (or don't care) I think break statements should also be added here:
- In gb_interface_read_and_clear_init_status() for the default case
- In gb_interface_activate() for the default case.
- In gb_svc_process_deferred_request() for the default case

But let's wait to see what Johan (et al) says. If you don't want to do that, say so and I'll do it later.

I looked at the code in drivers/staging/greybus/ and saw no need to add a "fallthrough" anywhere, but:
- In fw_mgmt_backend_fw_version_operation() Viresh seems to have skipped the break in the fault statement
- In gb_uart_request_handler() Bryan did this too.

Depending on discussion, these could be fixed in a separate patch as well.

These cases aren't found by "checkpatch.pl" because it only looks at case "blocks" that are followed by another case. So the last case isn't checked.

-Alex

Quoted 33 lines
---
  drivers/greybus/es2.c       | 2 +-
  drivers/greybus/interface.c | 2 +-
  2 files changed, 2 insertions(+), 2 deletions(-)
drivers/greybus/es2.c +1 -1
index 366716f11b1a..1df6ab5d339d 100644
@@ -759,7 +759,7 @@ static int check_urb_status(struct urb *urb)
  	case -EOVERFLOW:
  		dev_err(dev, "%s: overflow actual length is %d\n",
  			__func__, urb->actual_length);
-		/* fall through */
+		fallthrough;
  	case -ECONNRESET:
  	case -ENOENT:
  	case -ESHUTDOWN:
drivers/greybus/interface.c +1 -1
index 67dbe6fda9a1..58ea374d8aaa 100644
@@ -1233,7 +1233,7 @@ int gb_interface_add(struct gb_interface *intf)
  	case GB_INTERFACE_TYPE_GREYBUS:
  		dev_info(&intf->dev, "GMP VID=0x%08x, PID=0x%08x\n",
  			 intf->vendor_id, intf->product_id);
-		/* fall-through */
+		fallthrough;
  	case GB_INTERFACE_TYPE_UNIPRO:
  		dev_info(&intf->dev, "DDBL1 Manufacturer=0x%08x, Product=0x%08x\n",
  			 intf->ddbl1_manufacturer_id,
#2 lore raw
Quoted 6 lines · Alex Elder
Quoted 4 lines · Gustavo A. R. Silva

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Quoted 8 lines

Thanks for the patch. It looks good, but it raises another question I'd like discussion on.

It seems that Johan likes default (or final) cases in switch statements without a "break" statement. Viresh and Bryan appear to be fond of this too.

It's pedantic, but I don't like that. Am I wrong?

No, you are not wrong.

#3 lore raw
In reply to Alex Elder
Quoted 9 lines · Alex Elder
Quoted 4 lines · Gustavo A. R. Silva

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Thanks for the patch. It looks good, but it raises another question I'd like discussion on.

Quoted 6 lines

It seems that Johan likes default (or final) cases in switch statements without a "break" statement. Viresh and Bryan appear to be fond of this too.

It's pedantic, but I don't like that. Am I wrong?
--> Johan/Viresh/Bryan would you please comment?

I am not fond of them as they aren't required for the working of the code. It is a bit like using an empty return statement for a routine with void return type, though it surely adds some consistency to the switch case.

But if people really feel it must be there, then its fine :)

Signature · 1 line
viresh
#4 lore raw
Quoted 14 lines · Viresh Kumar
Quoted 13 lines · Alex Elder
Quoted 4 lines · Gustavo A. R. Silva

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Thanks for the patch. It looks good, but it raises another question I'd like discussion on.

It seems that Johan likes default (or final) cases in switch statements without a "break" statement. Viresh and Bryan appear to be fond of this too.

Quoted 6 lines
Quoted 2 lines

It's pedantic, but I don't like that. Am I wrong?
--> Johan/Viresh/Bryan would you please comment?

I am not fond of them as they aren't required for the working of the code. It is a bit like using an empty return statement for a routine with void return type, though it surely adds some consistency to the switch case.

I understand this perspective, and it's exactly why I wanted to have a conversation about it (rather than just saying it should be fixed). As similar example, I don't like unnecessary parentheses, but sometimes it's a good idea to have them.

Thanks.

-Alex

Quoted 2 lines

But if people really feel it must be there, then its fine :)

Reviewed-by
#5 lore raw
In reply to Alex Elder
Quoted 8 lines · Alex Elder
Quoted 4 lines · Gustavo A. R. Silva

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Thanks for the patch.  It looks good, but it raises another question I'd like discussion on.

It's been a week, and we heard back from Viresh (and Joe) on this, but no one else. Viresh left out the break statement on the last case of the switch statement intentionally, arguing that it is not needed (much like a return statement at the end of a void function). But he doesn't feel strongly enough insist it should stay that way. I'm sure the others omitted the break statement intentionally as well.

Given no strong pushback, I'll ask you (Gustavo) to post a second patch adding the missing break statements I described (and look for any others I might have missed). If you would prefer not to do that, just say so, and I will send out such a patch myself.

On your original patch, it looks good to me. Thank you.

Quoted 67 lines

It seems that Johan likes default (or final) cases in switch statements without a "break" statement.  Viresh and Bryan appear to be fond of this too.

It's pedantic, but I don't like that.  Am I wrong? --> Johan/Viresh/Bryan would you please comment?

If they aren't convincing (or don't care) I think break statements should also be added here:
- In gb_interface_read_and_clear_init_status() for the default case
- In gb_interface_activate() for the default case.
- In gb_svc_process_deferred_request() for the default case

But let's wait to see what Johan (et al) says.  If you don't want to do that, say so and I'll do it later.

I looked at the code in drivers/staging/greybus/ and saw no need to add a "fallthrough" anywhere, but:
- In fw_mgmt_backend_fw_version_operation() Viresh seems to have skipped the break in the fault statement
- In gb_uart_request_handler() Bryan did this too.

Depending on discussion, these could be fixed in a separate patch as well.

These cases aren't found by "checkpatch.pl" because it only looks at case "blocks" that are followed by another case. So the last case isn't checked.

                    -Alex

Quoted 33 lines

---
  drivers/greybus/es2.c       | 2 +-
  drivers/greybus/interface.c | 2 +-

  2 files changed, 2 insertions(+), 2 deletions(-)
drivers/greybus/es2.c +0 -0
index 366716f11b1a..1df6ab5d339d 100644
@@ -759,7 +759,7 @@ static int check_urb_status(struct urb *urb)

      case -EOVERFLOW:
          dev_err(dev, "%s: overflow actual length is %d\n", __func__, urb->actual_length);
-        /* fall through */
+        fallthrough;
      case -ECONNRESET:
      case -ENOENT:
      case -ESHUTDOWN:

drivers/greybus/interface.c +0 -0
index 67dbe6fda9a1..58ea374d8aaa 100644
@@ -1233,7 +1233,7 @@ int gb_interface_add(struct gb_interface *intf)

      case GB_INTERFACE_TYPE_GREYBUS:
          dev_info(&intf->dev, "GMP VID=0x%08x, PID=0x%08x\n", intf->vendor_id, intf->product_id);
-        /* fall-through */
+        fallthrough;
      case GB_INTERFACE_TYPE_UNIPRO:
          dev_info(&intf->dev, "DDBL1 Manufacturer=0x%08x, Product=0x%08x\n", intf->ddbl1_manufacturer_id,

maintainer
#6 lore raw
Quoted 9 lines · Alex Elder
Quoted 8 lines · Alex Elder
Quoted 4 lines · Gustavo A. R. Silva

Replace the existing /* fall through */ comments and its variants with the new pseudo-keyword macro fallthrough[1].

[1] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Thanks for the patch.  It looks good, but it raises another question I'd like discussion on.

Sorry about the late reply on this.

Quoted 7 lines

It's been a week, and we heard back from Viresh (and Joe) on this, but no one else. Viresh left out the break statement on the last case of the switch statement intentionally, arguing that it is not needed (much like a return statement at the end of a void function). But he doesn't feel strongly enough insist it should stay that way. I'm sure the others omitted the break statement intentionally as well.

I really don't mind break statements in the final case and often do add them, but I'm a bit reluctant to suggest that this is something that need "fixing". There are a ton of these all over the kernel, and I think we have too many opportunities for people to do mechanical clean ups already.

Especially after Gustavo's work, the only real argument for adding them is mostly moot as the compiler would catch it if anyone adds a new case and forgets about the break statement.

Quoted 5 lines

Given no strong pushback, I'll ask you (Gustavo) to post a second patch adding the missing break statements I described (and look for any others I might have missed). If you would prefer not to do that, just say so, and I will send out such a patch myself.

I'd probably just leave it as is, if only to not inspire others to send copy-cat "fixes".

Johan