nvme: multipath: round-robin fixes
cover nvme: multipath: round-robin fixes
From: Martin Wilck <[email protected]>
Hello Christoph, Sagi, Hannes,
The version of the patch set "nvme-multipath: fixes for non-optimized paths" that was eventually merged fails for the case where just a single, non-optimized path is available.
This mini-series fixes it, and simplifies the code somewhat.
Regards,
Martin
Martin Wilck (2):
nvme: multipath: round-robin: fix single non-optimized path case
nvme: multipath: round-robin: eliminate "fallback" variable
drivers/nvme/host/multipath.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-)
Signature · 7 lines
2.28.0 _______________________________________________ Linux-nvme mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-nvme
PATCH 1/2 nvme: multipath: round-robin: fix single non-optimized path case
From: Martin Wilck <[email protected]>
If there's only one usable, non-optimized path, nvme_round_robin_path() returns NULL, which is wrong. Fix it by falling back to "old", like in the single optimized path case. Also, if the active path isn't changed, there's no need to re-assign the pointer.
--- drivers/nvme/host/multipath.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-)
drivers/nvme/host/multipath.c +10 -5
@@ -255,12 +255,17 @@ static struct nvme_ns *nvme_round_robin_path(struct nvme_ns_head *head, fallback = ns; } -/* No optimized path found, re-check the current path */ +/* +* The loop above skips the current path for round-robin semantics. +* Fall back to the current path if either: +* - no other optimized path found and current is optimized, +* - no other usable path found and current is usable. +*/ if (!nvme_path_is_disabled (old) && - old->ana_state ==NVME_ANA_OPTIMIZED ) { - found = old; - goto out; - } + (old->ana_state ==NVME_ANA_OPTIMIZED || + (!fallback && old->ana_state ==NVME_ANA_NONOPTIMIZED ))) +return old; +if (!fallback)return NULL ; found = fallback;
Signature · 7 lines
2.28.0 _______________________________________________ Linux-nvme mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-nvme
Re: [PATCH 1/2] nvme: multipath: round-robin: fix single non-optimized path case
Quoted 10 lines · On Thu, 2020-08-06 at 15:19 +0200, [email protected] wrote:
From: Martin Wilck <[email protected]>
If there's only one usable, non-optimized path,
nvme_round_robin_path()
returns NULL, which is wrong. Fix it by falling back to "old", like in
the single optimized path case. Also, if the active path isn't changed,
there's no need to re-assign the pointer.
Quoted 4 lines
Fixes 3f6e3246db0e ("nvme-multipath: fix logic for non-optimizedpaths")
Signed-off-by Martin Wilck <[email protected]>Reported-by Martin George <[email protected]>
I made a mistake here. It should be
because Martin G. provided the initial idea for the fix.
_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
George, Martin
Tested-by Martin GeorgeRe: [PATCH 1/2] nvme: multipath: round-robin: fix single non-optimized path case
Quoted 16 lines · Martin Wilck
Quoted 10 lines · On Thu, 2020-08-06 at 15:19 +0200, [email protected] wrote:
From: Martin Wilck <[email protected]>
If there's only one usable, non-optimized path,
nvme_round_robin_path()
returns NULL, which is wrong. Fix it by falling back to "old", like in
the single optimized path case. Also, if the active path isn't changed,
there's no need to re-assign the pointer.Quoted 4 lines
Fixes 3f6e3246db0e ("nvme-multipath: fix logic for non-optimizedpaths")
Signed-off-by Martin Wilck <[email protected]>Reported-by Martin George <[email protected]>
Quoted 6 lines
I made a mistake here. It should be
Signed-off-by Martin George <[email protected]>because Martin G. provided the initial idea for the fix.
Thanks MartinW.
I've also tested this patch and verified that it works fine for all scenarios including the single ANA non-optimized path scenario. So in addition to MartinW's Signed-off-by tag, please add
as well for this patch. Thank you.
-Martin George
_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
PATCH 2/2 nvme: multipath: round-robin: eliminate "fallback" variable
In reply to [email protected]From: Martin Wilck <[email protected]>
If we find an optimized path, we quit the loop immediately. Thus we can use just one variable for the next path, slighly simplifying the code.
--- drivers/nvme/host/multipath.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-)
drivers/nvme/host/multipath.c +4 -5
@@ -233,7 +233,7 @@ static struct nvme_ns *nvme_next_ns(struct nvme_ns_head *head,static struct nvme_ns *nvme_round_robin_path (struct nvme_ns_head *head,int node,struct nvme_ns *old) { -struct nvme_ns *ns, *found, *fallback =NULL ; +struct nvme_ns *ns, *found =NULL ;if (list_is_singular (&head->list)) {if (nvme_path_is_disabled (old))
@@ -252,7 +252,7 @@ static struct nvme_ns *nvme_round_robin_path(struct nvme_ns_head *head, goto out; }if (ns->ana_state ==NVME_ANA_NONOPTIMIZED ) - fallback = ns; + found = ns; }/*
@@ -263,12 +263,11 @@ static struct nvme_ns *nvme_round_robin_path(struct nvme_ns_head *head,*/ if (!nvme_path_is_disabled (old) && (old->ana_state ==NVME_ANA_OPTIMIZED || - (!fallback && old->ana_state ==NVME_ANA_NONOPTIMIZED ))) + (!found && old->ana_state ==NVME_ANA_NONOPTIMIZED )))return old; -if (!fallback) +if (!found)return NULL ; - found = fallback; out:rcu_assign_pointer (head->current_path[node], found);return found;
Signature · 7 lines
2.28.0 _______________________________________________ Linux-nvme mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-nvme
Sagi Grimberg
maintainerThese look fine to me.
Can I get another set of eyes on this please?
_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
Quoted 3 lines · Sagi Grimberg
These look fine to me.
Can I get another set of eyes on this please?
Looks correct to me.
_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme