[PATCH] nvmet: nul-terminate the NQNs passed in the connect command

[PATCH] nvmet: nul-terminate the NQNs passed in the connect command

[Christoph Hellwig]

The host and subsystem NQNs are passed in the connect command payload and
interpreted as nul-terminated strings.  Ensure they actually are
nul-terminated before using them.

Fixes: a07b4970f464 "nvmet: add a generic NVMe target")
Reported-by: Alon Zahavi <zahavi.alon@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/nvme/target/fabrics-cmd.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/nvme/target/fabrics-cmd.c b/drivers/nvme/target/fabrics-cmd.c
index 43b5bd8bb6a52d..0920fe7ce4ac99 100644
--- a/drivers/nvme/target/fabrics-cmd.c
+++ b/drivers/nvme/target/fabrics-cmd.c
@@ -244,6 +244,8 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)
 		goto out;
 	}
 
+	d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';
+	d->hostnqn[NVMF_NQN_FIELD_LEN] = '\0';
 	status = nvmet_alloc_ctrl(d->subsysnqn, d->hostnqn, req,
 				  le32_to_cpu(c->kato), &ctrl);
 	if (status)
@@ -313,6 +315,8 @@ static void nvmet_execute_io_connect(struct nvmet_req *req)
 		goto out;
 	}
 
+	d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';
+	d->hostnqn[NVMF_NQN_FIELD_LEN] = '\0';
 	ctrl = nvmet_ctrl_find_get(d->subsysnqn, d->hostnqn,
 				   le16_to_cpu(d->cntlid), req);
 	if (!ctrl) {
-- 
2.39.2

Re: [PATCH] nvmet: nul-terminate the NQNs passed in the connect command

[Caleb Sander]

Won't this overrun the subsysnqn and hostnqn fields? Writing to
subsysnqn[NVMF_NQN_FIELD_LEN] will clobber the first char of hostnqn:
struct nvmf_connect_data {
        uuid_t hostid;
        __le16 cntlid;
        char resv4[238];
        char subsysnqn[NVMF_NQN_FIELD_LEN];
        char hostnqn[NVMF_NQN_FIELD_LEN];
        char resv5[256];
};

I think clearing the previous byte (index NVMF_NQN_FIELD_LEN - 1)
would work. The spec requires NQNs to be under 223 bytes anyways, so
they should never take up the whole field.

On Thu, Nov 9, 2023 at 10:45 PM Christoph Hellwig <hch@lst.de> wrote:
>
> The host and subsystem NQNs are passed in the connect command payload and
> interpreted as nul-terminated strings.  Ensure they actually are
> nul-terminated before using them.
>
> Fixes: a07b4970f464 "nvmet: add a generic NVMe target")
> Reported-by: Alon Zahavi <zahavi.alon@gmail.com>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  drivers/nvme/target/fabrics-cmd.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/drivers/nvme/target/fabrics-cmd.c b/drivers/nvme/target/fabrics-cmd.c
> index 43b5bd8bb6a52d..0920fe7ce4ac99 100644
> --- a/drivers/nvme/target/fabrics-cmd.c
> +++ b/drivers/nvme/target/fabrics-cmd.c
> @@ -244,6 +244,8 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)
>                 goto out;
>         }
>
> +       d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';
> +       d->hostnqn[NVMF_NQN_FIELD_LEN] = '\0';
>         status = nvmet_alloc_ctrl(d->subsysnqn, d->hostnqn, req,
>                                   le32_to_cpu(c->kato), &ctrl);
>         if (status)
> @@ -313,6 +315,8 @@ static void nvmet_execute_io_connect(struct nvmet_req *req)
>                 goto out;
>         }
>
> +       d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';
> +       d->hostnqn[NVMF_NQN_FIELD_LEN] = '\0';
>         ctrl = nvmet_ctrl_find_get(d->subsysnqn, d->hostnqn,
>                                    le16_to_cpu(d->cntlid), req);
>         if (!ctrl) {
> --
> 2.39.2
>
>

Re: [PATCH] nvmet: nul-terminate the NQNs passed in the connect command

[Alon Zahavi]

On Fri, 10 Nov 2023 at 17:37, Caleb Sander <csander@purestorage.com> wrote:
>
> Won't this overrun the subsysnqn and hostnqn fields? Writing to
> subsysnqn[NVMF_NQN_FIELD_LEN] will clobber the first char of hostnqn:
> struct nvmf_connect_data {
>         uuid_t hostid;
>         __le16 cntlid;
>         char resv4[238];
>         char subsysnqn[NVMF_NQN_FIELD_LEN];
>         char hostnqn[NVMF_NQN_FIELD_LEN];
>         char resv5[256];
> };
>
> I think clearing the previous byte (index NVMF_NQN_FIELD_LEN - 1)
> would work. The spec requires NQNs to be under 223 bytes anyways, so
> they should never take up the whole field.
>

True.
We should zero the byte in the `NVMF_NQN_SIZE` index, making the
fields their real size.
Changing the patch from `d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';` to
`d->subsysnqn[NVMF_NQN_SIZE] = '\0';`

> On Thu, Nov 9, 2023 at 10:45 PM Christoph Hellwig <hch@lst.de> wrote:
> >
> > The host and subsystem NQNs are passed in the connect command payload and
> > interpreted as nul-terminated strings.  Ensure they actually are
> > nul-terminated before using them.
> >
> > Fixes: a07b4970f464 "nvmet: add a generic NVMe target")
> > Reported-by: Alon Zahavi <zahavi.alon@gmail.com>
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> >  drivers/nvme/target/fabrics-cmd.c | 4 ++++
> >  1 file changed, 4 insertions(+)
> >
> > diff --git a/drivers/nvme/target/fabrics-cmd.c b/drivers/nvme/target/fabrics-cmd.c
> > index 43b5bd8bb6a52d..0920fe7ce4ac99 100644
> > --- a/drivers/nvme/target/fabrics-cmd.c
> > +++ b/drivers/nvme/target/fabrics-cmd.c
> > @@ -244,6 +244,8 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)
> >                 goto out;
> >         }
> >
> > +       d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';
> > +       d->hostnqn[NVMF_NQN_FIELD_LEN] = '\0';
> >         status = nvmet_alloc_ctrl(d->subsysnqn, d->hostnqn, req,
> >                                   le32_to_cpu(c->kato), &ctrl);
> >         if (status)
> > @@ -313,6 +315,8 @@ static void nvmet_execute_io_connect(struct nvmet_req *req)
> >                 goto out;
> >         }
> >
> > +       d->subsysnqn[NVMF_NQN_FIELD_LEN] = '\0';
> > +       d->hostnqn[NVMF_NQN_FIELD_LEN] = '\0';
> >         ctrl = nvmet_ctrl_find_get(d->subsysnqn, d->hostnqn,
> >                                    le16_to_cpu(d->cntlid), req);
> >         if (!ctrl) {
> > --
> > 2.39.2
> >
> >