2021-09-03 08:45:43

by Li Zhijian

[permalink] [raw]
Subject: [PATCH] IB/mlx5: unify return value to ENOENT

Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
the errors all occur at get_prefetchable_mr().

flags = IBV_ADVISE_MR_FLAG_FLUSH:
mlx5_ib_advise_mr_prefetch()
-> mlx5_ib_prefetch_sg_list()
-> get_prefetchable_mr()
return -ENOENT;

flags = 0:
mlx5_ib_advise_mr_prefetch()
-> init_prefetch_work()
-> get_prefetchable_mr()
return -EINVAL;

Signed-off-by: Li Zhijian <[email protected]>
---
drivers/infiniband/hw/mlx5/odp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index d0d98e584ebc..52572e7ea6f6 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -1828,7 +1828,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,

if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
destroy_prefetch_work(work);
- return -EINVAL;
+ return -ENOENT;
}
queue_work(system_unbound_wq, &work->work);
return 0;
--
2.31.1




2021-09-28 17:13:28

by Jason Gunthorpe

[permalink] [raw]
Subject: Re: [PATCH] IB/mlx5: unify return value to ENOENT

On Fri, Sep 03, 2021 at 04:48:15PM +0800, Li Zhijian wrote:
> Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
> the errors all occur at get_prefetchable_mr().

What do you think about this instead?

From b739920ed4869decb02a0dbc58256e6c72ec7061 Mon Sep 17 00:00:00 2001
From: Jason Gunthorpe <[email protected]>
Date: Fri, 3 Sep 2021 16:48:15 +0800
Subject: [PATCH] IB/mlx5: Flow through a more detailed return code from
get_prefetchable_mr()

The error returns for various cases detected by get_prefetchable_mr() get
confused as it flows back to userspace. Properly label each error path and
flow the error code properly back to the system call.

Suggested-by: Li Zhijian <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>
---
drivers/infiniband/hw/mlx5/odp.c | 40 ++++++++++++++++++++------------
1 file changed, 25 insertions(+), 15 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index d0d98e584ebcc3..77890a85fc2dd3 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -1708,20 +1708,26 @@ get_prefetchable_mr(struct ib_pd *pd, enum ib_uverbs_advise_mr_advice advice,

xa_lock(&dev->odp_mkeys);
mmkey = xa_load(&dev->odp_mkeys, mlx5_base_mkey(lkey));
- if (!mmkey || mmkey->key != lkey || mmkey->type != MLX5_MKEY_MR)
+ if (!mmkey || mmkey->key != lkey) {
+ mr = ERR_PTR(-ENOENT);
goto end;
+ }
+ if (mmkey->type != MLX5_MKEY_MR) {
+ mr = ERR_PTR(-EINVAL);
+ goto end;
+ }

mr = container_of(mmkey, struct mlx5_ib_mr, mmkey);

if (mr->ibmr.pd != pd) {
- mr = NULL;
+ mr = ERR_PTR(-EPERM);
goto end;
}

/* prefetch with write-access must be supported by the MR */
if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE &&
!mr->umem->writable) {
- mr = NULL;
+ mr = ERR_PTR(-EPERM);
goto end;
}

@@ -1753,7 +1759,7 @@ static void mlx5_ib_prefetch_mr_work(struct work_struct *w)
destroy_prefetch_work(work);
}

-static bool init_prefetch_work(struct ib_pd *pd,
+static int init_prefetch_work(struct ib_pd *pd,
enum ib_uverbs_advise_mr_advice advice,
u32 pf_flags, struct prefetch_mr_work *work,
struct ib_sge *sg_list, u32 num_sge)
@@ -1764,17 +1770,19 @@ static bool init_prefetch_work(struct ib_pd *pd,
work->pf_flags = pf_flags;

for (i = 0; i < num_sge; ++i) {
+ struct mlx5_ib_mr *mr;
+
+ mr = get_prefetchable_mr(pd, advice, sg_list[i].lkey);
+ if (IS_ERR(mr)) {
+ work->num_sge = i;
+ return PTR_ERR(mr);
+ }
work->frags[i].io_virt = sg_list[i].addr;
work->frags[i].length = sg_list[i].length;
- work->frags[i].mr =
- get_prefetchable_mr(pd, advice, sg_list[i].lkey);
- if (!work->frags[i].mr) {
- work->num_sge = i;
- return false;
- }
+ work->frags[i].mr = mr;
}
work->num_sge = num_sge;
- return true;
+ return 0;
}

static int mlx5_ib_prefetch_sg_list(struct ib_pd *pd,
@@ -1790,8 +1798,8 @@ static int mlx5_ib_prefetch_sg_list(struct ib_pd *pd,
struct mlx5_ib_mr *mr;

mr = get_prefetchable_mr(pd, advice, sg_list[i].lkey);
- if (!mr)
- return -ENOENT;
+ if (IS_ERR(mr))
+ return PTR_ERR(mr);
ret = pagefault_mr(mr, sg_list[i].addr, sg_list[i].length,
&bytes_mapped, pf_flags);
if (ret < 0) {
@@ -1811,6 +1819,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
{
u32 pf_flags = 0;
struct prefetch_mr_work *work;
+ int rc;

if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH)
pf_flags |= MLX5_PF_FLAGS_DOWNGRADE;
@@ -1826,9 +1835,10 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
if (!work)
return -ENOMEM;

- if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
+ rc = init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge);
+ if (rc) {
destroy_prefetch_work(work);
- return -EINVAL;
+ return rc;
}
queue_work(system_unbound_wq, &work->work);
return 0;
--
2.33.0

2021-09-29 04:31:30

by Zhijian Li (Fujitsu)

[permalink] [raw]
Subject: Re: [PATCH] IB/mlx5: unify return value to ENOENT

ping


On 03/09/2021 16:48, Li Zhijian wrote:
> Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
> the errors all occur at get_prefetchable_mr().
>
> flags = IBV_ADVISE_MR_FLAG_FLUSH:
> mlx5_ib_advise_mr_prefetch()
> -> mlx5_ib_prefetch_sg_list()
> -> get_prefetchable_mr()
> return -ENOENT;
>
> flags = 0:
> mlx5_ib_advise_mr_prefetch()
> -> init_prefetch_work()
> -> get_prefetchable_mr()
> return -EINVAL;
>
> Signed-off-by: Li Zhijian <[email protected]>
> ---
> drivers/infiniband/hw/mlx5/odp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
> index d0d98e584ebc..52572e7ea6f6 100644
> --- a/drivers/infiniband/hw/mlx5/odp.c
> +++ b/drivers/infiniband/hw/mlx5/odp.c
> @@ -1828,7 +1828,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
>
> if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
> destroy_prefetch_work(work);
> - return -EINVAL;
> + return -ENOENT;
> }
> queue_work(system_unbound_wq, &work->work);
> return 0;

2021-09-29 04:48:44

by Zhijian Li (Fujitsu)

[permalink] [raw]
Subject: Re: [PATCH] IB/mlx5: unify return value to ENOENT

So sorry, I missed this reply at previous ping


On 29/09/2021 01:08, Jason Gunthorpe wrote:
> On Fri, Sep 03, 2021 at 04:48:15PM +0800, Li Zhijian wrote:
>> Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
>> the errors all occur at get_prefetchable_mr().
> What do you think about this instead?
Thank you, it's much better.

Thanks
Zhijian




>
> From b739920ed4869decb02a0dbc58256e6c72ec7061 Mon Sep 17 00:00:00 2001
> From: Jason Gunthorpe <[email protected]>
> Date: Fri, 3 Sep 2021 16:48:15 +0800
> Subject: [PATCH] IB/mlx5: Flow through a more detailed return code from
> get_prefetchable_mr()
>
> The error returns for various cases detected by get_prefetchable_mr() get
> confused as it flows back to userspace. Properly label each error path and
> flow the error code properly back to the system call.
>
> Suggested-by: Li Zhijian <[email protected]>
> Signed-off-by: Jason Gunthorpe <[email protected]>
> ---
> drivers/infiniband/hw/mlx5/odp.c | 40 ++++++++++++++++++++------------
> 1 file changed, 25 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
> index d0d98e584ebcc3..77890a85fc2dd3 100644
> --- a/drivers/infiniband/hw/mlx5/odp.c
> +++ b/drivers/infiniband/hw/mlx5/odp.c
> @@ -1708,20 +1708,26 @@ get_prefetchable_mr(struct ib_pd *pd, enum ib_uverbs_advise_mr_advice advice,
>
> xa_lock(&dev->odp_mkeys);
> mmkey = xa_load(&dev->odp_mkeys, mlx5_base_mkey(lkey));
> - if (!mmkey || mmkey->key != lkey || mmkey->type != MLX5_MKEY_MR)
> + if (!mmkey || mmkey->key != lkey) {
> + mr = ERR_PTR(-ENOENT);
> goto end;
> + }
> + if (mmkey->type != MLX5_MKEY_MR) {
> + mr = ERR_PTR(-EINVAL);
> + goto end;
> + }
>
> mr = container_of(mmkey, struct mlx5_ib_mr, mmkey);
>
> if (mr->ibmr.pd != pd) {
> - mr = NULL;
> + mr = ERR_PTR(-EPERM);
> goto end;
> }
>
> /* prefetch with write-access must be supported by the MR */
> if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE &&
> !mr->umem->writable) {
> - mr = NULL;
> + mr = ERR_PTR(-EPERM);
> goto end;
> }
>
> @@ -1753,7 +1759,7 @@ static void mlx5_ib_prefetch_mr_work(struct work_struct *w)
> destroy_prefetch_work(work);
> }
>
> -static bool init_prefetch_work(struct ib_pd *pd,
> +static int init_prefetch_work(struct ib_pd *pd,
> enum ib_uverbs_advise_mr_advice advice,
> u32 pf_flags, struct prefetch_mr_work *work,
> struct ib_sge *sg_list, u32 num_sge)
> @@ -1764,17 +1770,19 @@ static bool init_prefetch_work(struct ib_pd *pd,
> work->pf_flags = pf_flags;
>
> for (i = 0; i < num_sge; ++i) {
> + struct mlx5_ib_mr *mr;
> +
> + mr = get_prefetchable_mr(pd, advice, sg_list[i].lkey);
> + if (IS_ERR(mr)) {
> + work->num_sge = i;
> + return PTR_ERR(mr);
> + }
> work->frags[i].io_virt = sg_list[i].addr;
> work->frags[i].length = sg_list[i].length;
> - work->frags[i].mr =
> - get_prefetchable_mr(pd, advice, sg_list[i].lkey);
> - if (!work->frags[i].mr) {
> - work->num_sge = i;
> - return false;
> - }
> + work->frags[i].mr = mr;
> }
> work->num_sge = num_sge;
> - return true;
> + return 0;
> }
>
> static int mlx5_ib_prefetch_sg_list(struct ib_pd *pd,
> @@ -1790,8 +1798,8 @@ static int mlx5_ib_prefetch_sg_list(struct ib_pd *pd,
> struct mlx5_ib_mr *mr;
>
> mr = get_prefetchable_mr(pd, advice, sg_list[i].lkey);
> - if (!mr)
> - return -ENOENT;
> + if (IS_ERR(mr))
> + return PTR_ERR(mr);
> ret = pagefault_mr(mr, sg_list[i].addr, sg_list[i].length,
> &bytes_mapped, pf_flags);
> if (ret < 0) {
> @@ -1811,6 +1819,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
> {
> u32 pf_flags = 0;
> struct prefetch_mr_work *work;
> + int rc;
>
> if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH)
> pf_flags |= MLX5_PF_FLAGS_DOWNGRADE;
> @@ -1826,9 +1835,10 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
> if (!work)
> return -ENOMEM;
>
> - if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
> + rc = init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge);
> + if (rc) {
> destroy_prefetch_work(work);
> - return -EINVAL;
> + return rc;
> }
> queue_work(system_unbound_wq, &work->work);
> return 0;

2021-10-01 14:50:43

by Jason Gunthorpe

[permalink] [raw]
Subject: Re: [PATCH] IB/mlx5: unify return value to ENOENT

On Wed, Sep 29, 2021 at 04:37:22AM +0000, [email protected] wrote:
> So sorry, I missed this reply at previous ping
>
>
> On 29/09/2021 01:08, Jason Gunthorpe wrote:
> > On Fri, Sep 03, 2021 at 04:48:15PM +0800, Li Zhijian wrote:
> >> Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
> >> the errors all occur at get_prefetchable_mr().
> > What do you think about this instead?
> Thank you, it's much better.

Ok, applied to for-next

Jason

2021-10-13 07:28:28

by Zhijian Li (Fujitsu)

[permalink] [raw]
Subject: Re: [PATCH] IB/mlx5: unify return value to ENOENT

Hi Jason

When update the ibv_advise_mr man page, i have a few concerns:


On 29/09/2021 01:08, Jason Gunthorpe wrote:
> On Fri, Sep 03, 2021 at 04:48:15PM +0800, Li Zhijian wrote:
>> Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
>> the errors all occur at get_prefetchable_mr().
> What do you think about this instead?
>
> From b739920ed4869decb02a0dbc58256e6c72ec7061 Mon Sep 17 00:00:00 2001
> From: Jason Gunthorpe <[email protected]>
> Date: Fri, 3 Sep 2021 16:48:15 +0800
> Subject: [PATCH] IB/mlx5: Flow through a more detailed return code from
> get_prefetchable_mr()
>
> The error returns for various cases detected by get_prefetchable_mr() get
> confused as it flows back to userspace. Properly label each error path and
> flow the error code properly back to the system call.
>
> Suggested-by: Li Zhijian <[email protected]>
> Signed-off-by: Jason Gunthorpe <[email protected]>
> ---
> drivers/infiniband/hw/mlx5/odp.c | 40 ++++++++++++++++++++------------
> 1 file changed, 25 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
> index d0d98e584ebcc3..77890a85fc2dd3 100644
> --- a/drivers/infiniband/hw/mlx5/odp.c
> +++ b/drivers/infiniband/hw/mlx5/odp.c
> @@ -1708,20 +1708,26 @@ get_prefetchable_mr(struct ib_pd *pd, enum ib_uverbs_advise_mr_advice advice,
>
> xa_lock(&dev->odp_mkeys);
> mmkey = xa_load(&dev->odp_mkeys, mlx5_base_mkey(lkey));
> - if (!mmkey || mmkey->key != lkey || mmkey->type != MLX5_MKEY_MR)
> + if (!mmkey || mmkey->key != lkey) {
> + mr = ERR_PTR(-ENOENT);
> goto end;
> + }
> + if (mmkey->type != MLX5_MKEY_MR) {
> + mr = ERR_PTR(-EINVAL);
> + goto end;
> + }


Can we return EINVAL in both above 2 cases so that we can attribute them to *lkey is invalid*  simply.
Otherwise it's hard to describe 2nd case by man page since users/developers cannot link mmkey->type
to the parameters of ibv_advise_mr().


>
> mr = container_of(mmkey, struct mlx5_ib_mr, mmkey);
>
> if (mr->ibmr.pd != pd) {
> - mr = NULL;
> + mr = ERR_PTR(-EPERM);

EINVAL is better  for compatible ? since man page said EINVAL in this case before.

Thanks


> goto end;
> }
>
> /* prefetch with write-access must be supported by the MR */
> if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE &&
> !mr->umem->writable) {
> - mr = NULL;
> + mr = ERR_PTR(-EPERM);
> goto end;
> }
>
> @@ -1753,7 +1759,7 @@ static void mlx5_ib_prefetch_mr_work(struct work_struct *w)
> destroy_prefetch_work(work);
> }
>
> -static bool init_prefetch_work(struct ib_pd *pd,
> +static int init_prefetch_work(struct ib_pd *pd,
> enum ib_uverbs_advise_mr_advice advice,
> u32 pf_flags, struct prefetch_mr_work *work,
> struct ib_sge *sg_list, u32 num_sge)
> @@ -1764,17 +1770,19 @@ static bool init_prefetch_work(struct ib_pd *pd,
> work->pf_flags = pf_flags;
>
> for (i = 0; i < num_sge; ++i) {
> + struct mlx5_ib_mr *mr;
> +
> + mr = get_prefetchable_mr(pd, advice, sg_list[i].lkey);
> + if (IS_ERR(mr)) {
> + work->num_sge = i;
> + return PTR_ERR(mr);
> + }
> work->frags[i].io_virt = sg_list[i].addr;
> work->frags[i].length = sg_list[i].length;
> - work->frags[i].mr =
> - get_prefetchable_mr(pd, advice, sg_list[i].lkey);
> - if (!work->frags[i].mr) {
> - work->num_sge = i;
> - return false;
> - }
> + work->frags[i].mr = mr;
> }
> work->num_sge = num_sge;
> - return true;
> + return 0;
> }
>
> static int mlx5_ib_prefetch_sg_list(struct ib_pd *pd,
> @@ -1790,8 +1798,8 @@ static int mlx5_ib_prefetch_sg_list(struct ib_pd *pd,
> struct mlx5_ib_mr *mr;
>
> mr = get_prefetchable_mr(pd, advice, sg_list[i].lkey);
> - if (!mr)
> - return -ENOENT;
> + if (IS_ERR(mr))
> + return PTR_ERR(mr);
> ret = pagefault_mr(mr, sg_list[i].addr, sg_list[i].length,
> &bytes_mapped, pf_flags);
> if (ret < 0) {
> @@ -1811,6 +1819,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
> {
> u32 pf_flags = 0;
> struct prefetch_mr_work *work;
> + int rc;
>
> if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH)
> pf_flags |= MLX5_PF_FLAGS_DOWNGRADE;
> @@ -1826,9 +1835,10 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
> if (!work)
> return -ENOMEM;
>
> - if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
> + rc = init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge);
> + if (rc) {
> destroy_prefetch_work(work);
> - return -EINVAL;
> + return rc;
> }
> queue_work(system_unbound_wq, &work->work);
> return 0;

2021-10-13 18:05:41

by Jason Gunthorpe

[permalink] [raw]
Subject: Re: [PATCH] IB/mlx5: unify return value to ENOENT

On Wed, Oct 13, 2021 at 07:26:49AM +0000, [email protected] wrote:
> Hi Jason
>
> When update the ibv_advise_mr man page, i have a few concerns:
>
>
> On 29/09/2021 01:08, Jason Gunthorpe wrote:
> > On Fri, Sep 03, 2021 at 04:48:15PM +0800, Li Zhijian wrote:
> >> Previously, ENOENT or EINVAL will be returned by ibv_advise_mr() although
> >> the errors all occur at get_prefetchable_mr().
> > What do you think about this instead?
> >
> > From b739920ed4869decb02a0dbc58256e6c72ec7061 Mon Sep 17 00:00:00 2001
> > From: Jason Gunthorpe <[email protected]>
> > Date: Fri, 3 Sep 2021 16:48:15 +0800
> > Subject: [PATCH] IB/mlx5: Flow through a more detailed return code from
> > get_prefetchable_mr()
> >
> > The error returns for various cases detected by get_prefetchable_mr() get
> > confused as it flows back to userspace. Properly label each error path and
> > flow the error code properly back to the system call.
> >
> > Suggested-by: Li Zhijian <[email protected]>
> > Signed-off-by: Jason Gunthorpe <[email protected]>
> > drivers/infiniband/hw/mlx5/odp.c | 40 ++++++++++++++++++++------------
> > 1 file changed, 25 insertions(+), 15 deletions(-)
> >
> > diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
> > index d0d98e584ebcc3..77890a85fc2dd3 100644
> > +++ b/drivers/infiniband/hw/mlx5/odp.c
> > @@ -1708,20 +1708,26 @@ get_prefetchable_mr(struct ib_pd *pd, enum ib_uverbs_advise_mr_advice advice,
> >
> > xa_lock(&dev->odp_mkeys);
> > mmkey = xa_load(&dev->odp_mkeys, mlx5_base_mkey(lkey));
> > - if (!mmkey || mmkey->key != lkey || mmkey->type != MLX5_MKEY_MR)
> > + if (!mmkey || mmkey->key != lkey) {
> > + mr = ERR_PTR(-ENOENT);
> > goto end;
> > + }
> > + if (mmkey->type != MLX5_MKEY_MR) {
> > + mr = ERR_PTR(-EINVAL);
> > + goto end;
> > + }
>
>
> Can we return EINVAL in both above 2 cases so that we can attribute
> them to *lkey is invalid*  simply. Otherwise it's hard to describe
> 2nd case by man page since users/developers cannot link mmkey->type
> to the parameters of ibv_advise_mr().

kley is valid in the 2nd case, but points to the wrong kidn of object
to prefetch, hence EIVNAL. Eg it is a MW or something.

> > mr = container_of(mmkey, struct mlx5_ib_mr, mmkey);
> >
> > if (mr->ibmr.pd != pd) {
> > - mr = NULL;
> > + mr = ERR_PTR(-EPERM);
>
> EINVAL is better  for compatible ? since man page said EINVAL in this case before.

Referencing a valid lkey outside the caller's security scope should be
EPERM.

Jason