Mailbox: Add support for Platform Communication Channel

Hello Ashwin Chaugule,

The patch 86c22f8c9a3b: "Mailbox: Add support for Platform Communication Channel" from Nov 12, 2014, leads to the following static checker warning:

drivers/mailbox/pcc.c:119 pcc_mbox_request_channel()
error: 'chan' dereferencing possible ERR_PTR()

drivers/mailbox/pcc.c
110 /*

   111           * Each PCC Subspace is a Mailbox Channel.
   112           * The PCC Clients get their PCC Subspace ID
   113           * from their own tables and pass it here.
   114           * This returns a pointer to the PCC subspace
   115           * for the Client to operate on.
   116           */
   117          chan = get_pcc_channel(subspace_id);
   118  
   119          if (!chan || chan->cl) {
                    ^^^^^

Should be checking IS_ERR(). Also is the error message correct?

   120                  dev_err(dev, "%s: PCC mailbox not free\n", __func__);
   121                  return ERR_PTR(-EBUSY);
   122          }

	drivers/mailbox/pcc.c:220 pcc_send_data()
	warn: impossible condition '(ss_idx < 0) => (0-65535 < 0)'

   215          u16 cmd = *(u16 *) data;
   216          u16 ss_idx = -1;
                ^^^^^^^^^^^^^^^^

Make this an int. Remove the bogus initializer.

   217  
   218          ss_idx = get_subspace_id(chan);
   219  

Don't put a blank line between the function calls and the error handling, they are part of the same paragraph.

   220          if (ss_idx < 0) {
   221                  pr_err("Invalid Subspace ID from PCC client\n");
   222                  return -EINVAL;
   223          }
   224  

Also this file has some Sparse warnings:

drivers/mailbox/pcc.c:103:18: warning: symbol 'pcc_mbox_request_channel' was not declared. Should it be static? drivers/mailbox/pcc.c:146:6: warning: symbol 'pcc_mbox_free_channel' was not declared. Should it be static? drivers/mailbox/pcc.c:180:18: warning: incorrect type in argument 1 (different address spaces)
drivers/mailbox/pcc.c:180:18: expected void const volatile [noderef] <asn:2>*addr
drivers/mailbox/pcc.c:180:18: got unsigned short *<noident>
drivers/mailbox/pcc.c:230:22: warning: incorrect type in argument 2 (different address spaces)
drivers/mailbox/pcc.c:230:22: expected void volatile [noderef] <asn:2>*addr
drivers/mailbox/pcc.c:230:22: got unsigned short *<noident>
drivers/mailbox/pcc.c:233:47: warning: incorrect type in argument 2 (different address spaces)
drivers/mailbox/pcc.c:233:47: expected void volatile [noderef] <asn:2>*addr
drivers/mailbox/pcc.c:233:47: got unsigned int *<noident>
drivers/mailbox/pcc.c:236:20: warning: incorrect type in argument 2 (different address spaces)
drivers/mailbox/pcc.c:236:20: expected void volatile [noderef] <asn:2>*addr
drivers/mailbox/pcc.c:236:20: got unsigned short *<noident>
drivers/mailbox/pcc.c:369:24: warning: symbol 'pcc_mbox_driver' was not declared. Should it be static?

regards,
dan carpenter

Hi Dan,

Will fix.

Sure.

Ok.

I've looked at these previously too and I believe they're false positives. Will send a follow up patch at the earliest.

Thanks,
Ashwin

They're not bugs, it just means that there are some __iomem annotations missing and some functions should be made static or declared in a .h file or whatever.

Something like this:

index 6dbf6fc..ccf0b18 100644
@@ -171,8 +171,8 @@ EXPORT_SYMBOL_GPL(pcc_mbox_free_channel);
 static bool pcc_tx_done(struct mbox_chan *chan)
 {
 	struct acpi_pcct_hw_reduced *pcct_ss = chan->con_priv;
-	struct acpi_pcct_shared_memory *generic_comm_base -		(struct acpi_pcct_shared_memory *) pcct_ss->base_address;
+	struct acpi_pcct_shared_memory __iomem *generic_comm_base +		(struct acpi_pcct_shared_memory __iomem *) pcct_ss->base_address;
 	u16 cmd_delay = pcct_ss->latency;
 	unsigned int retries = 0;
 

@@ -206,8 +206,8 @@ static bool pcc_tx_done(struct mbox_chan *chan)
 static int pcc_send_data(struct mbox_chan *chan, void *data)
 {
 	struct acpi_pcct_hw_reduced *pcct_ss = chan->con_priv;
-	struct acpi_pcct_shared_memory *generic_comm_base -		(struct acpi_pcct_shared_memory *) pcct_ss->base_address;
+	struct acpi_pcct_shared_memory __iomem *generic_comm_base +		(struct acpi_pcct_shared_memory __iomem *) pcct_ss->base_address;
 	struct acpi_generic_address doorbell;
 	u64 doorbell_preserve;
 	u64 doorbell_val;
 

Hi Dan,

True. We've been testing this driver on a non X86 platform in the past few days and had to make a few modifications which make some of these changes go away. Will send out a patch after some more testing. Very much appreciate your review!

Cheers,
Ashwin