'status' has been overwritten to 0 after nfsd4_ssc_setup_dul(), this
cause 0 will be return in vfs_kern_mount() error case. Fix to return
nfserr_nodev in this error.
Fixes: f4e44b393389 ("NFSD: delay unmount source's export after inter-server copy completed.")
Reported-by: Hulk Robot <[email protected]>
Signed-off-by: Wei Yongjun <[email protected]>
---
v1 -> v2: change to return nfserr_nodev
---
fs/nfsd/nfs4proc.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 0bd71c6da81d..b082cbde3e07 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1323,6 +1323,7 @@ nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
module_put(type->owner);
if (IS_ERR(ss_mnt)) {
+ status = nfserr_nodev;
if (work)
nfsd4_ssc_cancel_dul_work(nn, work);
goto out_free_devname;
On Fri, Jun 04, 2021 at 10:12:37AM +0000, Wei Yongjun wrote:
> 'status' has been overwritten to 0 after nfsd4_ssc_setup_dul(), this
> cause 0 will be return in vfs_kern_mount() error case. Fix to return
> nfserr_nodev in this error.
Why is that the right error? I don't see it mentioned among the errors
COPY can return:
https://datatracker.ietf.org/doc/html/rfc7862#page-50
It might be reasonable to just map the error returned by vfs_kern_mount
to an nfs error, I don't know--we'd need to think about the different
errors vfs_kern_mount might return.
OFFLOAD_DENIED seems safe as it should cause a fallback to a normal
copy.
Same goes for the other spot here where we return nodev.
--b.
>
> Fixes: f4e44b393389 ("NFSD: delay unmount source's export after inter-server copy completed.")
> Reported-by: Hulk Robot <[email protected]>
> Signed-off-by: Wei Yongjun <[email protected]>
> ---
> v1 -> v2: change to return nfserr_nodev
> ---
> fs/nfsd/nfs4proc.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index 0bd71c6da81d..b082cbde3e07 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -1323,6 +1323,7 @@ nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
> ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
> module_put(type->owner);
> if (IS_ERR(ss_mnt)) {
> + status = nfserr_nodev;
> if (work)
> nfsd4_ssc_cancel_dul_work(nn, work);
> goto out_free_devname;
On 6/4/21 7:40 AM, J. Bruce Fields wrote:
> On Fri, Jun 04, 2021 at 10:12:37AM +0000, Wei Yongjun wrote:
>> 'status' has been overwritten to 0 after nfsd4_ssc_setup_dul(), this
>> cause 0 will be return in vfs_kern_mount() error case. Fix to return
>> nfserr_nodev in this error.
> Why is that the right error?
That was the original error before it was overwritten by
nfsd4_ssc_setup_dul. However, it actually does not matter which
error nfsd4_interssc_connect returns to caller since nfsd4_copy
maps all errors to nfserr_offload_denied before returning it
to the client.
-Dai
> I don't see it mentioned among the errors
> COPY can return:
>
> https://urldefense.com/v3/__https://datatracker.ietf.org/doc/html/rfc7862*page-50__;Iw!!GqivPVa7Brio!NSHjr5li4vkJl0VAWWz2mrW2wiVn9wcJfK7ZJON3bmM8REVUqI82bS4FFQdajQ$
>
> It might be reasonable to just map the error returned by vfs_kern_mount
> to an nfs error, I don't know--we'd need to think about the different
> errors vfs_kern_mount might return.
>
> OFFLOAD_DENIED seems safe as it should cause a fallback to a normal
> copy.
>
> Same goes for the other spot here where we return nodev.
>
> --b.
>
>> Fixes: f4e44b393389 ("NFSD: delay unmount source's export after inter-server copy completed.")
>> Reported-by: Hulk Robot <[email protected]>
>> Signed-off-by: Wei Yongjun <[email protected]>
>> ---
>> v1 -> v2: change to return nfserr_nodev
>> ---
>> fs/nfsd/nfs4proc.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
>> index 0bd71c6da81d..b082cbde3e07 100644
>> --- a/fs/nfsd/nfs4proc.c
>> +++ b/fs/nfsd/nfs4proc.c
>> @@ -1323,6 +1323,7 @@ nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
>> ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
>> module_put(type->owner);
>> if (IS_ERR(ss_mnt)) {
>> + status = nfserr_nodev;
>> if (work)
>> nfsd4_ssc_cancel_dul_work(nn, work);
>> goto out_free_devname;
On Fri, Jun 04, 2021 at 09:36:58AM -0700, [email protected] wrote:
>
> On 6/4/21 7:40 AM, J. Bruce Fields wrote:
> >On Fri, Jun 04, 2021 at 10:12:37AM +0000, Wei Yongjun wrote:
> >>'status' has been overwritten to 0 after nfsd4_ssc_setup_dul(), this
> >>cause 0 will be return in vfs_kern_mount() error case. Fix to return
> >>nfserr_nodev in this error.
> >Why is that the right error?
>
> That was the original error before it was overwritten by
> nfsd4_ssc_setup_dul. However, it actually does not matter which
> error nfsd4_interssc_connect returns to caller since nfsd4_copy
> maps all errors to nfserr_offload_denied before returning it
> to the client.
OK, I overlooked that.--b.
>
> -Dai
>
> >I don't see it mentioned among the errors
> >COPY can return:
> >
> > https://urldefense.com/v3/__https://datatracker.ietf.org/doc/html/rfc7862*page-50__;Iw!!GqivPVa7Brio!NSHjr5li4vkJl0VAWWz2mrW2wiVn9wcJfK7ZJON3bmM8REVUqI82bS4FFQdajQ$
> >
> >It might be reasonable to just map the error returned by vfs_kern_mount
> >to an nfs error, I don't know--we'd need to think about the different
> >errors vfs_kern_mount might return.
> >
> >OFFLOAD_DENIED seems safe as it should cause a fallback to a normal
> >copy.
> >
> >Same goes for the other spot here where we return nodev.
> >
> >--b.
> >
> >>Fixes: f4e44b393389 ("NFSD: delay unmount source's export after inter-server copy completed.")
> >>Reported-by: Hulk Robot <[email protected]>
> >>Signed-off-by: Wei Yongjun <[email protected]>
> >>---
> >>v1 -> v2: change to return nfserr_nodev
> >>---
> >> fs/nfsd/nfs4proc.c | 1 +
> >> 1 file changed, 1 insertion(+)
> >>
> >>diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> >>index 0bd71c6da81d..b082cbde3e07 100644
> >>--- a/fs/nfsd/nfs4proc.c
> >>+++ b/fs/nfsd/nfs4proc.c
> >>@@ -1323,6 +1323,7 @@ nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
> >> ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
> >> module_put(type->owner);
> >> if (IS_ERR(ss_mnt)) {
> >>+ status = nfserr_nodev;
> >> if (work)
> >> nfsd4_ssc_cancel_dul_work(nn, work);
> >> goto out_free_devname;