[PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration

[PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration

[Shiraz Saleem]

From: Christopher Bednarz <christopher.n.bednarz@intel.com>

Currently irdma allows zero-length STAGs to be programmed in HW during
the kernel mode fast register flow. Zero-length MR or STAG registration
disable HW memory length checks.

Improve gaps in bounds checking in irdma by preventing zero-length STAG or
MR registrations except if the IB_PD_UNSAFE_GLOBAL_RKEY is set.

This addresses the disclosure CVE-2023-25775.

Fixes: b48c24c2d710 ("RDMA/irdma: Implement device supported verb APIs")
Signed-off-by: Christopher Bednarz <christopher.n.bednarz@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
---
 drivers/infiniband/hw/irdma/ctrl.c  |  6 ++++++
 drivers/infiniband/hw/irdma/type.h  |  2 ++
 drivers/infiniband/hw/irdma/verbs.c | 10 ++++++++--
 3 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/irdma/ctrl.c b/drivers/infiniband/hw/irdma/ctrl.c
index d88c9184007e..2bc340deb6bb 100644
--- a/drivers/infiniband/hw/irdma/ctrl.c
+++ b/drivers/infiniband/hw/irdma/ctrl.c
@@ -1061,6 +1061,9 @@ static int irdma_sc_alloc_stag(struct irdma_sc_dev *dev,
 	u64 hdr;
 	enum irdma_page_size page_size;
 
+	if (!info->total_len && !info->all_memory)
+		return -EINVAL;
+
 	if (info->page_size == 0x40000000)
 		page_size = IRDMA_PAGE_SIZE_1G;
 	else if (info->page_size == 0x200000)
@@ -1126,6 +1129,9 @@ static int irdma_sc_mr_reg_non_shared(struct irdma_sc_dev *dev,
 	u8 addr_type;
 	enum irdma_page_size page_size;
 
+	if (!info->total_len && !info->all_memory)
+		return -EINVAL;
+
 	if (info->page_size == 0x40000000)
 		page_size = IRDMA_PAGE_SIZE_1G;
 	else if (info->page_size == 0x200000)
diff --git a/drivers/infiniband/hw/irdma/type.h b/drivers/infiniband/hw/irdma/type.h
index 5ee68604e59f..16ada4c2ced0 100644
--- a/drivers/infiniband/hw/irdma/type.h
+++ b/drivers/infiniband/hw/irdma/type.h
@@ -969,6 +969,7 @@ struct irdma_allocate_stag_info {
 	bool remote_access:1;
 	bool use_hmc_fcn_index:1;
 	bool use_pf_rid:1;
+	bool all_memory:1;
 	u8 hmc_fcn_index;
 };
 
@@ -996,6 +997,7 @@ struct irdma_reg_ns_stag_info {
 	bool use_hmc_fcn_index:1;
 	u8 hmc_fcn_index;
 	bool use_pf_rid:1;
+	bool all_memory:1;
 };
 
 struct irdma_fast_reg_stag_info {
diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
index a7b82aea4d08..4110cc159bd9 100644
--- a/drivers/infiniband/hw/irdma/verbs.c
+++ b/drivers/infiniband/hw/irdma/verbs.c
@@ -2576,7 +2576,8 @@ static int irdma_hw_alloc_stag(struct irdma_device *iwdev,
 			       struct irdma_mr *iwmr)
 {
 	struct irdma_allocate_stag_info *info;
-	struct irdma_pd *iwpd = to_iwpd(iwmr->ibmr.pd);
+	struct ib_pd *pd = iwmr->ibmr.pd;
+	struct irdma_pd *iwpd = to_iwpd(pd);
 	int status;
 	struct irdma_cqp_request *cqp_request;
 	struct cqp_cmds_info *cqp_info;
@@ -2592,6 +2593,7 @@ static int irdma_hw_alloc_stag(struct irdma_device *iwdev,
 	info->stag_idx = iwmr->stag >> IRDMA_CQPSQ_STAG_IDX_S;
 	info->pd_id = iwpd->sc_pd.pd_id;
 	info->total_len = iwmr->len;
+	info->all_memory = pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY;
 	info->remote_access = true;
 	cqp_info->cqp_cmd = IRDMA_OP_ALLOC_STAG;
 	cqp_info->post_sq = 1;
@@ -2639,6 +2641,8 @@ static struct ib_mr *irdma_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
 	iwmr->type = IRDMA_MEMREG_TYPE_MEM;
 	palloc = &iwpbl->pble_alloc;
 	iwmr->page_cnt = max_num_sg;
+	/* Use system PAGE_SIZE as the sg page sizes are unknown at this point */
+	iwmr->len = max_num_sg * PAGE_SIZE;
 	err_code = irdma_get_pble(iwdev->rf->pble_rsrc, palloc, iwmr->page_cnt,
 				  false);
 	if (err_code)
@@ -2718,7 +2722,8 @@ static int irdma_hwreg_mr(struct irdma_device *iwdev, struct irdma_mr *iwmr,
 {
 	struct irdma_pbl *iwpbl = &iwmr->iwpbl;
 	struct irdma_reg_ns_stag_info *stag_info;
-	struct irdma_pd *iwpd = to_iwpd(iwmr->ibmr.pd);
+	struct ib_pd *pd = iwmr->ibmr.pd;
+	struct irdma_pd *iwpd = to_iwpd(pd);
 	struct irdma_pble_alloc *palloc = &iwpbl->pble_alloc;
 	struct irdma_cqp_request *cqp_request;
 	struct cqp_cmds_info *cqp_info;
@@ -2737,6 +2742,7 @@ static int irdma_hwreg_mr(struct irdma_device *iwdev, struct irdma_mr *iwmr,
 	stag_info->total_len = iwmr->len;
 	stag_info->access_rights = irdma_get_mr_access(access);
 	stag_info->pd_id = iwpd->sc_pd.pd_id;
+	stag_info->all_memory = pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY;
 	if (stag_info->access_rights & IRDMA_ACCESS_FLAGS_ZERO_BASED)
 		stag_info->addr_type = IRDMA_ADDR_TYPE_ZERO_BASED;
 	else
-- 
1.8.3.1

Re: [PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration

[Leon Romanovsky]

On Fri, Aug 18, 2023 at 09:48:38AM -0500, Shiraz Saleem wrote:
> From: Christopher Bednarz <christopher.n.bednarz@intel.com>
> 
> Currently irdma allows zero-length STAGs to be programmed in HW during
> the kernel mode fast register flow. Zero-length MR or STAG registration
> disable HW memory length checks.
> 
> Improve gaps in bounds checking in irdma by preventing zero-length STAG or
> MR registrations except if the IB_PD_UNSAFE_GLOBAL_RKEY is set.
> 
> This addresses the disclosure CVE-2023-25775.
> 
> Fixes: b48c24c2d710 ("RDMA/irdma: Implement device supported verb APIs")
> Signed-off-by: Christopher Bednarz <christopher.n.bednarz@intel.com>
> Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
> ---
>  drivers/infiniband/hw/irdma/ctrl.c  |  6 ++++++
>  drivers/infiniband/hw/irdma/type.h  |  2 ++
>  drivers/infiniband/hw/irdma/verbs.c | 10 ++++++++--
>  3 files changed, 16 insertions(+), 2 deletions(-)

I applied it to rdma-next because we are in -rc6 now and anyway this
patch will land Linus very soon.

Thanks

Re: [PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration

[Leon Romanovsky]

On Fri, 18 Aug 2023 09:48:38 -0500, Shiraz Saleem wrote:
> Currently irdma allows zero-length STAGs to be programmed in HW during
> the kernel mode fast register flow. Zero-length MR or STAG registration
> disable HW memory length checks.
> 
> Improve gaps in bounds checking in irdma by preventing zero-length STAG or
> MR registrations except if the IB_PD_UNSAFE_GLOBAL_RKEY is set.
> 
> [...]

Applied, thanks!

[1/1] RDMA/irdma: Prevent zero-length STAG registration
      https://git.kernel.org/rdma/rdma/c/bb6d73d9add68a

Best regards,
-- 
Leon Romanovsky <leon@kernel.org>

Re: [PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration

[liujian (CE)]

On 2023/8/18 22:48, Shiraz Saleem wrote:
> From: Christopher Bednarz <christopher.n.bednarz@intel.com>
> 
> Currently irdma allows zero-length STAGs to be programmed in HW during
> the kernel mode fast register flow. Zero-length MR or STAG registration
> disable HW memory length checks.
> 
> Improve gaps in bounds checking in irdma by preventing zero-length STAG or
> MR registrations except if the IB_PD_UNSAFE_GLOBAL_RKEY is set.
> 
> This addresses the disclosure CVE-2023-25775.
> 
> Fixes: b48c24c2d710 ("RDMA/irdma: Implement device supported verb APIs")

Hello,I would like to consult the CVE. The driver corresponding to the 
kernel of an earlier version (< 5.14) is i40iw and has similar code 
logic. Is this CVE also involved?

> Signed-off-by: Christopher Bednarz <christopher.n.bednarz@intel.com>
> Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
> ---
>   drivers/infiniband/hw/irdma/ctrl.c  |  6 ++++++
>   drivers/infiniband/hw/irdma/type.h  |  2 ++
>   drivers/infiniband/hw/irdma/verbs.c | 10 ++++++++--
>   3 files changed, 16 insertions(+), 2 deletions(-)
> 
[...]

RE: [PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration

[Saleem, Shiraz]

> Subject: Re: [PATCH for-rc] RDMA/irdma: Prevent zero-length STAG registration
> 
> 
> 
> On 2023/8/18 22:48, Shiraz Saleem wrote:
> > From: Christopher Bednarz <christopher.n.bednarz@intel.com>
> >
> > Currently irdma allows zero-length STAGs to be programmed in HW during
> > the kernel mode fast register flow. Zero-length MR or STAG
> > registration disable HW memory length checks.
> >
> > Improve gaps in bounds checking in irdma by preventing zero-length
> > STAG or MR registrations except if the IB_PD_UNSAFE_GLOBAL_RKEY is
> set.
> >
> > This addresses the disclosure CVE-2023-25775.
> >
> > Fixes: b48c24c2d710 ("RDMA/irdma: Implement device supported verb
> > APIs")
> 
> Hello,I would like to consult the CVE. The driver corresponding to the kernel of
> an earlier version (< 5.14) is i40iw and has similar code logic. Is this CVE also
> involved?
> 

Yes. It’s a problem in i40iw too.

Shiraz