[ Apologies for sending this to all the usual suspects here ... ]
drivers/scsi/scsi_sysfs.c :: store_rescan_field() calls
scsi_rescan_device() without a prototype, and thus results in a compiler
warning.
Fix that up by adding the prototype to scsi.h, where it belongs.
But then we see we are storing the return value of a void function (so
that is why ANSI C is good)... so fix that up, too, by setting the
return value to zero if a valid device was found. Otherwise, return
ENODEV as before.
Patch is against 2.5.65.
Robert Love
drivers/scsi/scsi.h | 1 +
drivers/scsi/scsi_sysfs.c | 6 ++++--
2 files changed, 5 insertions(+), 2 deletions(-)
diff -urN linux-2.5.65/drivers/scsi/scsi.h linux/drivers/scsi/scsi.h
--- linux-2.5.65/drivers/scsi/scsi.h 2003-03-19 15:44:06.279618904 -0500
+++ linux/drivers/scsi/scsi.h 2003-03-19 15:39:47.355981280 -0500
@@ -443,6 +443,7 @@
extern int scsi_attach_device(struct scsi_device *);
extern void scsi_detach_device(struct scsi_device *);
extern int scsi_get_device_flags(unsigned char *vendor, unsigned char *model);
+extern void scsi_rescan_device(struct scsi_device *sdev);
/*
* Newer request-based interfaces.
diff -urN linux-2.5.65/drivers/scsi/scsi_sysfs.c linux/drivers/scsi/scsi_sysfs.c
--- linux-2.5.65/drivers/scsi/scsi_sysfs.c 2003-03-19 15:44:06.196631520 -0500
+++ linux/drivers/scsi/scsi_sysfs.c 2003-03-19 15:42:45.249937288 -0500
@@ -278,8 +278,10 @@
int ret = ENODEV;
struct scsi_device *sdev;
sdev = to_scsi_device(dev);
- if (sdev)
- ret = scsi_rescan_device(sdev);
+ if (sdev) {
+ ret = 0;
+ scsi_rescan_device(sdev);
+ }
return ret;
}
Robert Love wrote:
> drivers/scsi/scsi_sysfs.c :: store_rescan_field() calls
> scsi_rescan_device() without a prototype, and thus results
> in a compiler warning.
>
> Fix that up by adding the prototype to scsi.h, where it belongs.
> But then we see we are storing the return value of a void function
> (so that is why ANSI C is good)... so fix that up, too, by setting
> the return value to zero if a valid device was found. Otherwise,
> return ENODEV as before.
>
> Patch is against 2.5.65.
Christoph Hellwig sent a patch for this to the linux-scsi
list. Both patches have the same minor problem. Sysfs
store() callbacks are supposed to return the count of
bytes consumed. It's a moot point in this case since there
is no sscanf() processing. To use this code I can get
to the appropriate sysfs directory and write:
# echo "anything_I_like" > rescan
So are all bytes in that string consumed? If so the "ret = 0;"
line in your patch should be "ret = count;".
Doug Gilbert