Re: [PATCH v2 1/1] NFSD: fix problems with cleanup on errors in nfsd4_copy

[Chuck Lever III <chuck.lever@xxxxxxxxxx>]

> On Jan 31, 2023, at 8:36 AM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> 
> On Sat, 2023-01-28 at 20:44 -0800, Dai Ngo wrote:
>> When nfsd4_copy fails to allocate memory for async_copy->cp_src, or
>> nfs4_init_copy_state fails, it calls cleanup_async_copy to do the
>> cleanup for the async_copy which causes page fault since async_copy
>> is not yet initialized.
>> 
>> This patch rearranges the order of initializing the fields in
>> async_copy and adds checks in cleanup_async_copy to skip un-initialized
>> fields.
>> 
>> Fixes: ce0887ac96d3 ("NFSD add nfs4 inter ssc to nfsd4_copy")
>> Fixes: 87689df69491 ("NFSD: Shrink size of struct nfsd4_copy")
>> Signed-off-by: Dai Ngo <dai.ngo@xxxxxxxxxx>
>> ---
>> fs/nfsd/nfs4proc.c  | 12 ++++++++----
>> fs/nfsd/nfs4state.c |  5 +++--
>> 2 files changed, 11 insertions(+), 6 deletions(-)
>> 
>> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
>> index 57f791899de3..0754b38d3a43 100644
>> --- a/fs/nfsd/nfs4proc.c
>> +++ b/fs/nfsd/nfs4proc.c
>> @@ -1687,9 +1687,12 @@ static void cleanup_async_copy(struct nfsd4_copy *copy)
>> {
>> 	nfs4_free_copy_state(copy);
>> 	release_copy_files(copy);
>> -	spin_lock(&copy->cp_clp->async_lock);
>> -	list_del(&copy->copies);
>> -	spin_unlock(&copy->cp_clp->async_lock);
>> +	if (copy->cp_clp) {
>> +		spin_lock(&copy->cp_clp->async_lock);
>> +		if (!list_empty(&copy->copies))
>> +			list_del(&copy->copies);
> 
> Can we make this a list_del_init? If cleanup_async_copy were called on
> this twice, then the second time could end up corrupting the
> async_copies list. The cost difference is negligible here.

I noticed this yesterday and made the change in my tree.


>> +		spin_unlock(&copy->cp_clp->async_lock);
>> +	}
>> 	nfs4_put_copy(copy);
>> }
>> 
>> @@ -1786,12 +1789,13 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
>> 		async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
>> 		if (!async_copy)
>> 			goto out_err;
>> +		INIT_LIST_HEAD(&async_copy->copies);
>> +		refcount_set(&async_copy->refcount, 1);
>> 		async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
>> 		if (!async_copy->cp_src)
>> 			goto out_err;
>> 		if (!nfs4_init_copy_state(nn, copy))
>> 			goto out_err;
>> -		refcount_set(&async_copy->refcount, 1);
>> 		memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
>> 			sizeof(copy->cp_res.cb_stateid));
>> 		dup_copy_fields(copy, async_copy);
>> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
>> index ace02fd0d590..c39e43742dd6 100644
>> --- a/fs/nfsd/nfs4state.c
>> +++ b/fs/nfsd/nfs4state.c
>> @@ -975,7 +975,6 @@ static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
>> 
>> 	stid->cs_stid.si_opaque.so_clid.cl_boot = (u32)nn->boot_time;
>> 	stid->cs_stid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id;
>> -	stid->cs_type = cs_type;
>> 
>> 	idr_preload(GFP_KERNEL);
>> 	spin_lock(&nn->s2s_cp_lock);
>> @@ -986,6 +985,7 @@ static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
>> 	idr_preload_end();
>> 	if (new_id < 0)
>> 		return 0;
>> +	stid->cs_type = cs_type;
>> 	return 1;
>> }
>> 
>> @@ -1019,7 +1019,8 @@ void nfs4_free_copy_state(struct nfsd4_copy *copy)
>> {
>> 	struct nfsd_net *nn;
>> 
>> -	WARN_ON_ONCE(copy->cp_stateid.cs_type != NFS4_COPY_STID);
>> +	if (copy->cp_stateid.cs_type != NFS4_COPY_STID)
>> +		return;
> 
> It's probably fine to keep the WARN_ON_ONCE here. You're testing the
> condition anyway so you can do:
> 
>    if (WARN_ON_ONCE(copy->cp_stateid.cs_type != NFS4_COPY_STID))

Six of one...

I'm OK leaving the WARN out, it doesn't seem high value to me.


>> 	nn = net_generic(copy->cp_clp->net, nfsd_net_id);
>> 	spin_lock(&nn->s2s_cp_lock);
>> 	idr_remove(&nn->s2s_cp_stateids,
> 
> -- 
> Jeff Layton <jlayton@xxxxxxxxxx>

--
Chuck Lever