Thanks for your respond.
> 1) in parse_path() you need to always initialize q
the condition of q initialization is
if (m->retain_attached_hw_handler)”
comparing with previous condition
if (test_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags) || m->hw_handler_name)
it does not increase the possibility of q initialization.
> 2) setting m->hw_handler_name to attached_handler_name needs to still
> happen regardless of whether m->hw_handler_name was previously set
The patch does not change the logic above, does it?
> 3) the "retain_attached_hw_handler" feature should still be allowed on
> the command line, no sense to break multipath-tools
Yes, actually I am still preparing two other patches now,
one patch modifies parse_features() to ignore unsupported
parameter and do not return failure. Another patch modifies
multipath to remove this parameter also.
> In addition, your patch actually breaks the ability to cope with ?EBUSY
> from the call to scsi_dh_attach(). Meaning we wouldn't properly fall
> back to retaining the attached hw handler.
scsi_dh_attach() return ?EBUSY only if that device has attached
a different handle, after previous processing, the handle would
be same, so it is impossible to return ?EBUSY in scsi_dh_attach(),
we do not need to process it.
Regards,
Tang
发件人: Mike Snitzer <snitzer@xxxxxxxxxx>
收件人: tang.junhui@xxxxxxxxxx,
抄送: zhang.kai16@xxxxxxxxxx, dm-devel@xxxxxxxxxx, agk@xxxxxxxxxx
日期: 2016/11/29 05:52
主题: Re: [dm-devel] [PATCH 2/2] dm-mpath: Remove useless retain_attached_hw_handler parameter
发件人: dm-devel-bounces@xxxxxxxxxx
On Thu, Nov 24 2016 at 2:11am -0500,
tang.junhui@xxxxxxxxxx <tang.junhui@xxxxxxxxxx> wrote:
> From: "tang.junhui" <tang.junhui@xxxxxxxxxx>
>
> Hardware handle would be retained no matter parameter
> retain_attached_hw_handler is set or not in the logic
> of current code. So remove this useless parameter.
Right, that wasn't always the case. Previously (before commit )
dm-mpath would first detach the attached handler.
I'm not completely opposed to removing the code that checks
MPATHF_RETAIN_ATTACHED_HW_HANDLER in parse_path() but your proposed
patch is broken in 2 ways:
1) in parse_path() you need to always initialize q
2) setting m->hw_handler_name to attached_handler_name needs to still
happen regardless of whether m->hw_handler_name was previously set
3) the "retain_attached_hw_handler" feature should still be allowed on
the command line, no sense to break multipath-tools
But honestly, I'm not seeing any reason to not just leave the existing
code.
--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel
-- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
