On Fri, 2008-05-02 at 23:47 +0100, Alasdair G Kergon wrote:
> On Thu, May 01, 2008 at 02:50:17PM -0700, Chandra Seetharaman wrote:
> > Subject: [PATCH 06/10] scsi_dh: Do not allow arguments for hardware handlers
> > From: Chandra Seetharaman <sekharan@xxxxxxxxxx>
> > The new implementation does not use the arguments provided with the
> > hardware handler. This patch makes the change visible to the user by
> > failing table load when arguments are provided with the hardware handler.
>
> But then someone supplying such an arg may need to update their
> userspace multipath config file before they can use this new kernel.
>
> Past experience suggests that changes like that usually come back
> and bite us. Let's instead just have the kernel log a message noting
> that the args have been ignored and allow people to deal with it at
> their leisure.
Understand.
How does the following patch look ?
--------------------
Subject: [PATCH 06/10] scsi_dh: Print warning when arguments were provided for hardware handlers
From: Chandra Seetharaman <sekharan@xxxxxxxxxx>
The new implementation does not use the arguments provided with the
hardware handler. This patch makes the change visible to the user by
printing a warning to the kernel log.
Signed-off-by: Chandra Seetharaman <sekharan@xxxxxxxxxx>
---
Index: linux-2.6.25/drivers/md/dm-mpath.c
===================================================================
--- linux-2.6.25.orig/drivers/md/dm-mpath.c
+++ linux-2.6.25/drivers/md/dm-mpath.c
@@ -656,8 +656,11 @@ static int parse_hw_handler(struct arg_s
{0, 1024, "invalid number of hardware handler args"},
};
- if (read_param(_params, shift(as), &hw_argc, &ti->error))
- return -EINVAL;
+ if (read_param(_params, shift(as), &hw_argc, &ti->error)) {
+ if (hw_argc > 1)
+ printk(KERN_WARN "Additional arguments provided to "
+ "hardware handler were ignored\n");
+ }
if (!hw_argc)
return 0;
------------------------
>
> Alasdair
--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel
