Re: [PATCH] fs/dax: Fix pmd vs pte conflict detection
Dan Williams writes:
> Check for NULL entries before checking the entry order, otherwise NULL
> is misinterpreted as a present pte conflict. The 'order' check needs to
> happen before the locked check as an unlocked entry at the wrong order
> must fallback to lookup the correct order.
Please include the user-visible effects of the problem in the changelog.
Thanks,
Jeff
>
> Reported-by: Jeff Smits
> Reported-by: Doug Nelson
> Cc:
> Fixes: 23c84eb78375 ("dax: Fix missed wakeup with PMD faults")
> Cc: Jan Kara
> Cc: Matthew Wilcox (Oracle)
> Signed-off-by: Dan Williams
> ---
> fs/dax.c |5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/fs/dax.c b/fs/dax.c
> index a71881e77204..08160011d94c 100644
> --- a/fs/dax.c
> +++ b/fs/dax.c
> @@ -221,10 +221,11 @@ static void *get_unlocked_entry(struct xa_state *xas,
> unsigned int order)
>
> for (;;) {
> entry = xas_find_conflict(xas);
> + if (!entry || WARN_ON_ONCE(!xa_is_value(entry)))
> + return entry;
> if (dax_entry_order(entry) < order)
> return XA_RETRY_ENTRY;
> - if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) ||
> - !dax_is_locked(entry))
> + if (!dax_is_locked(entry))
> return entry;
>
> wq = dax_entry_waitqueue(xas, entry, );
___
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-le...@lists.01.org
Re: [PATCH] fs/dax: Fix pmd vs pte conflict detection
On Sat 19-10-19 09:26:19, Dan Williams wrote:
> Check for NULL entries before checking the entry order, otherwise NULL
> is misinterpreted as a present pte conflict. The 'order' check needs to
> happen before the locked check as an unlocked entry at the wrong order
> must fallback to lookup the correct order.
>
> Reported-by: Jeff Smits
> Reported-by: Doug Nelson
> Cc:
> Fixes: 23c84eb78375 ("dax: Fix missed wakeup with PMD faults")
> Cc: Jan Kara
> Cc: Matthew Wilcox (Oracle)
> Signed-off-by: Dan Williams
Good catch! The patch looks good to me. You can add:
Reviewed-by: Jan Kara
Honza
> ---
> fs/dax.c |5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/fs/dax.c b/fs/dax.c
> index a71881e77204..08160011d94c 100644
> --- a/fs/dax.c
> +++ b/fs/dax.c
> @@ -221,10 +221,11 @@ static void *get_unlocked_entry(struct xa_state *xas,
> unsigned int order)
>
> for (;;) {
> entry = xas_find_conflict(xas);
> + if (!entry || WARN_ON_ONCE(!xa_is_value(entry)))
> + return entry;
> if (dax_entry_order(entry) < order)
> return XA_RETRY_ENTRY;
> - if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) ||
> - !dax_is_locked(entry))
> + if (!dax_is_locked(entry))
> return entry;
>
> wq = dax_entry_waitqueue(xas, entry, );
>
--
Jan Kara
SUSE Labs, CR
___
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-le...@lists.01.org
Re: [PATCH] fs/dax: Fix pmd vs pte conflict detection
On Sat, Oct 19, 2019 at 4:09 PM Dan Williams wrote:
>
> On Sat, Oct 19, 2019 at 1:50 PM Matthew Wilcox wrote:
> >
> > On Sat, Oct 19, 2019 at 09:26:19AM -0700, Dan Williams wrote:
> > > Check for NULL entries before checking the entry order, otherwise NULL
> > > is misinterpreted as a present pte conflict. The 'order' check needs to
> > > happen before the locked check as an unlocked entry at the wrong order
> > > must fallback to lookup the correct order.
> > >
> > > Reported-by: Jeff Smits
> > > Reported-by: Doug Nelson
> > > Cc:
> > > Fixes: 23c84eb78375 ("dax: Fix missed wakeup with PMD faults")
> > > Cc: Jan Kara
> > > Cc: Matthew Wilcox (Oracle)
> > > Signed-off-by: Dan Williams
> > > ---
> > > fs/dax.c |5 +++--
> > > 1 file changed, 3 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/fs/dax.c b/fs/dax.c
> > > index a71881e77204..08160011d94c 100644
> > > --- a/fs/dax.c
> > > +++ b/fs/dax.c
> > > @@ -221,10 +221,11 @@ static void *get_unlocked_entry(struct xa_state
> > > *xas, unsigned int order)
> > >
> > > for (;;) {
> > > entry = xas_find_conflict(xas);
> > > + if (!entry || WARN_ON_ONCE(!xa_is_value(entry)))
> > > + return entry;
> > > if (dax_entry_order(entry) < order)
> > > return XA_RETRY_ENTRY;
> > > - if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) ||
> > > - !dax_is_locked(entry))
> > > + if (!dax_is_locked(entry))
> > > return entry;
> >
> > Yes, I think this works. Should we also add:
> >
> > static unsigned int dax_entry_order(void *entry)
> > {
> > + BUG_ON(!xa_is_value(entry));
> > if (xa_to_value(entry) & DAX_PMD)
> > return PMD_ORDER;
> > return 0;
> > }
> >
> > which would have caught this logic error before it caused a performance
> > regression?
>
> Sounds good will add it to v2.
...except that there are multiple dax helpers that have the 'value'
entry assumption. I'd rather do all of them in a separate patch, or
none of them. It turns out that after this change all
dax_entry_order() invocations are now protected by a xa_is_value()
assert earlier in the calling function.
___
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-le...@lists.01.org
Re: [PATCH] fs/dax: Fix pmd vs pte conflict detection
On Sat, Oct 19, 2019 at 1:50 PM Matthew Wilcox wrote:
>
> On Sat, Oct 19, 2019 at 09:26:19AM -0700, Dan Williams wrote:
> > Check for NULL entries before checking the entry order, otherwise NULL
> > is misinterpreted as a present pte conflict. The 'order' check needs to
> > happen before the locked check as an unlocked entry at the wrong order
> > must fallback to lookup the correct order.
> >
> > Reported-by: Jeff Smits
> > Reported-by: Doug Nelson
> > Cc:
> > Fixes: 23c84eb78375 ("dax: Fix missed wakeup with PMD faults")
> > Cc: Jan Kara
> > Cc: Matthew Wilcox (Oracle)
> > Signed-off-by: Dan Williams
> > ---
> > fs/dax.c |5 +++--
> > 1 file changed, 3 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/dax.c b/fs/dax.c
> > index a71881e77204..08160011d94c 100644
> > --- a/fs/dax.c
> > +++ b/fs/dax.c
> > @@ -221,10 +221,11 @@ static void *get_unlocked_entry(struct xa_state *xas,
> > unsigned int order)
> >
> > for (;;) {
> > entry = xas_find_conflict(xas);
> > + if (!entry || WARN_ON_ONCE(!xa_is_value(entry)))
> > + return entry;
> > if (dax_entry_order(entry) < order)
> > return XA_RETRY_ENTRY;
> > - if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) ||
> > - !dax_is_locked(entry))
> > + if (!dax_is_locked(entry))
> > return entry;
>
> Yes, I think this works. Should we also add:
>
> static unsigned int dax_entry_order(void *entry)
> {
> + BUG_ON(!xa_is_value(entry));
> if (xa_to_value(entry) & DAX_PMD)
> return PMD_ORDER;
> return 0;
> }
>
> which would have caught this logic error before it caused a performance
> regression?
Sounds good will add it to v2.
___
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-le...@lists.01.org
Re: [PATCH] fs/dax: Fix pmd vs pte conflict detection
On Sat, Oct 19, 2019 at 09:26:19AM -0700, Dan Williams wrote:
> Check for NULL entries before checking the entry order, otherwise NULL
> is misinterpreted as a present pte conflict. The 'order' check needs to
> happen before the locked check as an unlocked entry at the wrong order
> must fallback to lookup the correct order.
>
> Reported-by: Jeff Smits
> Reported-by: Doug Nelson
> Cc:
> Fixes: 23c84eb78375 ("dax: Fix missed wakeup with PMD faults")
> Cc: Jan Kara
> Cc: Matthew Wilcox (Oracle)
> Signed-off-by: Dan Williams
> ---
> fs/dax.c |5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/fs/dax.c b/fs/dax.c
> index a71881e77204..08160011d94c 100644
> --- a/fs/dax.c
> +++ b/fs/dax.c
> @@ -221,10 +221,11 @@ static void *get_unlocked_entry(struct xa_state *xas,
> unsigned int order)
>
> for (;;) {
> entry = xas_find_conflict(xas);
> + if (!entry || WARN_ON_ONCE(!xa_is_value(entry)))
> + return entry;
> if (dax_entry_order(entry) < order)
> return XA_RETRY_ENTRY;
> - if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) ||
> - !dax_is_locked(entry))
> + if (!dax_is_locked(entry))
> return entry;
Yes, I think this works. Should we also add:
static unsigned int dax_entry_order(void *entry)
{
+ BUG_ON(!xa_is_value(entry));
if (xa_to_value(entry) & DAX_PMD)
return PMD_ORDER;
return 0;
}
which would have caught this logic error before it caused a performance
regression?
___
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-le...@lists.01.org
[PATCH] fs/dax: Fix pmd vs pte conflict detection
Check for NULL entries before checking the entry order, otherwise NULL
is misinterpreted as a present pte conflict. The 'order' check needs to
happen before the locked check as an unlocked entry at the wrong order
must fallback to lookup the correct order.
Reported-by: Jeff Smits
Reported-by: Doug Nelson
Cc:
Fixes: 23c84eb78375 ("dax: Fix missed wakeup with PMD faults")
Cc: Jan Kara
Cc: Matthew Wilcox (Oracle)
Signed-off-by: Dan Williams
---
fs/dax.c |5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/fs/dax.c b/fs/dax.c
index a71881e77204..08160011d94c 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -221,10 +221,11 @@ static void *get_unlocked_entry(struct xa_state *xas,
unsigned int order)
for (;;) {
entry = xas_find_conflict(xas);
+ if (!entry || WARN_ON_ONCE(!xa_is_value(entry)))
+ return entry;
if (dax_entry_order(entry) < order)
return XA_RETRY_ENTRY;
- if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) ||
- !dax_is_locked(entry))
+ if (!dax_is_locked(entry))
return entry;
wq = dax_entry_waitqueue(xas, entry, );
___
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-le...@lists.01.org
