diff mbox series

[net,v2] octeontx2-af: Unlock contexts in the queue context cache in case of fault detection

Message ID 20230223110125.2172509-1-saikrishnag@marvell.com
State New
Headers
Series [net,v2] octeontx2-af: Unlock contexts in the queue context cache in case of fault detection |

Commit Message

Sai Krishna Gajula Feb. 23, 2023, 11:01 a.m. UTC 
  From: Suman Ghosh <sumang@marvell.com>

NDC caches contexts of frequently used queue's (Rx and Tx queues)
contexts. Due to a HW errata when NDC detects fault/poision while
accessing contexts it could go into an illegal state where a cache
line could get locked forever. To makesure all cache lines in NDC
are available for optimum performance upon fault/lockerror/posion
errors scan through all cache lines in NDC and clear the lock bit.

Fixes: 4a3581cd5995 ("octeontx2-af: NPA AQ instruction enqueue support")
Signed-off-by: Suman Ghosh <sumang@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
---
v2:
    - Rebased to latest net tree after net-next merge as suggested by Paolo

 .../net/ethernet/marvell/octeontx2/af/rvu.h   |  8 +++
 .../marvell/octeontx2/af/rvu_debugfs.c        |  3 -
 .../ethernet/marvell/octeontx2/af/rvu_nix.c   | 16 +++++-
 .../ethernet/marvell/octeontx2/af/rvu_npa.c   | 55 ++++++++++++++++++-
 .../ethernet/marvell/octeontx2/af/rvu_reg.h   |  3 +
 5 files changed, 80 insertions(+), 5 deletions(-)

Comments

Simon Horman Feb. 23, 2023, 1:16 p.m. UTC | #1 
On Thu, Feb 23, 2023 at 04:31:25PM +0530, Sai Krishna wrote:
> From: Suman Ghosh <sumang@marvell.com>
> 
> NDC caches contexts of frequently used queue's (Rx and Tx queues)
> contexts. Due to a HW errata when NDC detects fault/poision while
> accessing contexts it could go into an illegal state where a cache
> line could get locked forever. To makesure all cache lines in NDC
> are available for optimum performance upon fault/lockerror/posion
> errors scan through all cache lines in NDC and clear the lock bit.
> 
> Fixes: 4a3581cd5995 ("octeontx2-af: NPA AQ instruction enqueue support")
> Signed-off-by: Suman Ghosh <sumang@marvell.com>
> Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
> Signed-off-by: Sai Krishna <saikrishnag@marvell.com>

...

> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> index 389663a13d1d..6508f25b2b37 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> @@ -884,6 +884,12 @@ int rvu_cpt_lf_teardown(struct rvu *rvu, u16 pcifunc, int blkaddr, int lf,
>  int rvu_cpt_ctx_flush(struct rvu *rvu, u16 pcifunc);
>  int rvu_cpt_init(struct rvu *rvu);
>  
> +/* NDC APIs */
> +#define NDC_MAX_BANK(rvu, blk_addr) (rvu_read64(rvu, \
> +					blk_addr, NDC_AF_CONST) & 0xFF)
> +#define NDC_MAX_LINE_PER_BANK(rvu, blk_addr) ((rvu_read64(rvu, \
> +					blk_addr, NDC_AF_CONST) & 0xFFFF0000) >> 16)

Perhaps not appropriate to include as part of a fix,
as NDC_MAX_BANK is being moved from elsewhere,
but I wonder if this might be more cleanly implemented
using FIELD_GET().

...

> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> index 1729b22580ce..bc6ca5ccc1ff 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> @@ -694,6 +694,7 @@
>  #define NDC_AF_INTR_ENA_W1S		(0x00068)
>  #define NDC_AF_INTR_ENA_W1C		(0x00070)
>  #define NDC_AF_ACTIVE_PC		(0x00078)
> +#define NDC_AF_CAMS_RD_INTERVAL		(0x00080)
>  #define NDC_AF_BP_TEST_ENABLE		(0x001F8)
>  #define NDC_AF_BP_TEST(a)		(0x00200 | (a) << 3)
>  #define NDC_AF_BLK_RST			(0x002F0)
> @@ -709,6 +710,8 @@
>  		(0x00F00 | (a) << 5 | (b) << 4)
>  #define NDC_AF_BANKX_HIT_PC(a)		(0x01000 | (a) << 3)
>  #define NDC_AF_BANKX_MISS_PC(a)		(0x01100 | (a) << 3)
> +#define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> +		(0x10000 | (a) << 3 | (b) << 3)

It looks a little odd that both a and b are shifted by 3 bits.
If it's intended then perhaps it would be clearer to write this as:

#define NDC_AF_BANKX_LINEX_METADATA(a, b) \
		(0x10000 | ((a) | (b)) << 3)

>  
>  /* LBK */
>  #define LBK_CONST			(0x10ull)
> -- 
> 2.25.1
>
Sai Krishna Gajula Feb. 24, 2023, 8:39 a.m. UTC | #2 
Hi Simon,

> -----Original Message-----
> From: Simon Horman <simon.horman@corigine.com>
> Sent: Thursday, February 23, 2023 6:47 PM
> To: Sai Krishna Gajula <saikrishnag@marvell.com>
> Cc: davem@davemloft.net; edumazet@google.com; kuba@kernel.org;
> pabeni@redhat.com; netdev@vger.kernel.org; linux-kernel@vger.kernel.org;
> Sunil Kovvuri Goutham <sgoutham@marvell.com>; Suman Ghosh
> <sumang@marvell.com>
> Subject: Re: [net PATCH v2] octeontx2-af: Unlock contexts in the queue
> context cache in case of fault detection
> 
> 
> ----------------------------------------------------------------------
> On Thu, Feb 23, 2023 at 04:31:25PM +0530, Sai Krishna wrote:
> > From: Suman Ghosh <sumang@marvell.com>
> >
> > NDC caches contexts of frequently used queue's (Rx and Tx queues)
> > contexts. Due to a HW errata when NDC detects fault/poision while
> > accessing contexts it could go into an illegal state where a cache
> > line could get locked forever. To makesure all cache lines in NDC are
> > available for optimum performance upon fault/lockerror/posion errors
> > scan through all cache lines in NDC and clear the lock bit.
> >
> > Fixes: 4a3581cd5995 ("octeontx2-af: NPA AQ instruction enqueue
> > support")
> > Signed-off-by: Suman Ghosh <sumang@marvell.com>
> > Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
> > Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
> 
> ...
> 
> > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > index 389663a13d1d..6508f25b2b37 100644
> > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > @@ -884,6 +884,12 @@ int rvu_cpt_lf_teardown(struct rvu *rvu, u16
> > pcifunc, int blkaddr, int lf,  int rvu_cpt_ctx_flush(struct rvu *rvu,
> > u16 pcifunc);  int rvu_cpt_init(struct rvu *rvu);
> >
> > +/* NDC APIs */
> > +#define NDC_MAX_BANK(rvu, blk_addr) (rvu_read64(rvu, \
> > +					blk_addr, NDC_AF_CONST) & 0xFF)
> > +#define NDC_MAX_LINE_PER_BANK(rvu, blk_addr) ((rvu_read64(rvu, \
> > +					blk_addr, NDC_AF_CONST) &
> 0xFFFF0000) >> 16)
> 
> Perhaps not appropriate to include as part of a fix, as NDC_MAX_BANK is
> being moved from elsewhere, but I wonder if this might be more cleanly
> implemented using FIELD_GET().

We will modify and send a separate patch for all the possible macros that can be replaced by FIELD_GET(). 

> 
> ...
> 
> > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > index 1729b22580ce..bc6ca5ccc1ff 100644
> > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > @@ -694,6 +694,7 @@
> >  #define NDC_AF_INTR_ENA_W1S		(0x00068)
> >  #define NDC_AF_INTR_ENA_W1C		(0x00070)
> >  #define NDC_AF_ACTIVE_PC		(0x00078)
> > +#define NDC_AF_CAMS_RD_INTERVAL		(0x00080)
> >  #define NDC_AF_BP_TEST_ENABLE		(0x001F8)
> >  #define NDC_AF_BP_TEST(a)		(0x00200 | (a) << 3)
> >  #define NDC_AF_BLK_RST			(0x002F0)
> > @@ -709,6 +710,8 @@
> >  		(0x00F00 | (a) << 5 | (b) << 4)
> >  #define NDC_AF_BANKX_HIT_PC(a)		(0x01000 | (a) << 3)
> >  #define NDC_AF_BANKX_MISS_PC(a)		(0x01100 | (a) << 3)
> > +#define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> > +		(0x10000 | (a) << 3 | (b) << 3)
> 
> It looks a little odd that both a and b are shifted by 3 bits.
> If it's intended then perhaps it would be clearer to write this as:
> 
> #define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> 		(0x10000 | ((a) | (b)) << 3)

will send v3 patch.

Thanks,
Sai

> 
> >
> >  /* LBK */
> >  #define LBK_CONST			(0x10ull)
> > --
> > 2.25.1
> >
Simon Horman Feb. 24, 2023, 9:07 a.m. UTC | #3 
On Fri, Feb 24, 2023 at 08:39:20AM +0000, Sai Krishna Gajula wrote:
> Hi Simon,
> 
> > -----Original Message-----
> > From: Simon Horman <simon.horman@corigine.com>
> > Sent: Thursday, February 23, 2023 6:47 PM
> > To: Sai Krishna Gajula <saikrishnag@marvell.com>
> > Cc: davem@davemloft.net; edumazet@google.com; kuba@kernel.org;
> > pabeni@redhat.com; netdev@vger.kernel.org; linux-kernel@vger.kernel.org;
> > Sunil Kovvuri Goutham <sgoutham@marvell.com>; Suman Ghosh
> > <sumang@marvell.com>
> > Subject: Re: [net PATCH v2] octeontx2-af: Unlock contexts in the queue
> > context cache in case of fault detection
> > 
> > 
> > ----------------------------------------------------------------------
> > On Thu, Feb 23, 2023 at 04:31:25PM +0530, Sai Krishna wrote:
> > > From: Suman Ghosh <sumang@marvell.com>
> > >
> > > NDC caches contexts of frequently used queue's (Rx and Tx queues)
> > > contexts. Due to a HW errata when NDC detects fault/poision while
> > > accessing contexts it could go into an illegal state where a cache
> > > line could get locked forever. To makesure all cache lines in NDC are
> > > available for optimum performance upon fault/lockerror/posion errors
> > > scan through all cache lines in NDC and clear the lock bit.
> > >
> > > Fixes: 4a3581cd5995 ("octeontx2-af: NPA AQ instruction enqueue
> > > support")
> > > Signed-off-by: Suman Ghosh <sumang@marvell.com>
> > > Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
> > > Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
> > 
> > ...
> > 
> > > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > index 389663a13d1d..6508f25b2b37 100644
> > > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > @@ -884,6 +884,12 @@ int rvu_cpt_lf_teardown(struct rvu *rvu, u16
> > > pcifunc, int blkaddr, int lf,  int rvu_cpt_ctx_flush(struct rvu *rvu,
> > > u16 pcifunc);  int rvu_cpt_init(struct rvu *rvu);
> > >
> > > +/* NDC APIs */
> > > +#define NDC_MAX_BANK(rvu, blk_addr) (rvu_read64(rvu, \
> > > +					blk_addr, NDC_AF_CONST) & 0xFF)
> > > +#define NDC_MAX_LINE_PER_BANK(rvu, blk_addr) ((rvu_read64(rvu, \
> > > +					blk_addr, NDC_AF_CONST) &
> > 0xFFFF0000) >> 16)
> > 
> > Perhaps not appropriate to include as part of a fix, as NDC_MAX_BANK is
> > being moved from elsewhere, but I wonder if this might be more cleanly
> > implemented using FIELD_GET().
> 
> We will modify and send a separate patch for all the possible macros that can be replaced by FIELD_GET(). 

Thanks, much appreciated.

> > ...
> > 
> > > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > index 1729b22580ce..bc6ca5ccc1ff 100644
> > > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > @@ -694,6 +694,7 @@
> > >  #define NDC_AF_INTR_ENA_W1S		(0x00068)
> > >  #define NDC_AF_INTR_ENA_W1C		(0x00070)
> > >  #define NDC_AF_ACTIVE_PC		(0x00078)
> > > +#define NDC_AF_CAMS_RD_INTERVAL		(0x00080)
> > >  #define NDC_AF_BP_TEST_ENABLE		(0x001F8)
> > >  #define NDC_AF_BP_TEST(a)		(0x00200 | (a) << 3)
> > >  #define NDC_AF_BLK_RST			(0x002F0)
> > > @@ -709,6 +710,8 @@
> > >  		(0x00F00 | (a) << 5 | (b) << 4)
> > >  #define NDC_AF_BANKX_HIT_PC(a)		(0x01000 | (a) << 3)
> > >  #define NDC_AF_BANKX_MISS_PC(a)		(0x01100 | (a) << 3)
> > > +#define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> > > +		(0x10000 | (a) << 3 | (b) << 3)
> > 
> > It looks a little odd that both a and b are shifted by 3 bits.
> > If it's intended then perhaps it would be clearer to write this as:
> > 
> > #define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> > 		(0x10000 | ((a) | (b)) << 3)
> 
> will send v3 patch.

Likewise, thanks.
Sai Krishna Gajula March 2, 2023, 6:43 a.m. UTC | #4 
Hi Simon,

> -----Original Message-----
> From: Simon Horman <simon.horman@corigine.com>
> Sent: Friday, February 24, 2023 2:37 PM
> To: Sai Krishna Gajula <saikrishnag@marvell.com>
> Cc: davem@davemloft.net; edumazet@google.com; kuba@kernel.org;
> pabeni@redhat.com; netdev@vger.kernel.org; linux-kernel@vger.kernel.org;
> Sunil Kovvuri Goutham <sgoutham@marvell.com>; Suman Ghosh
> <sumang@marvell.com>
> Subject: Re: [net PATCH v2] octeontx2-af: Unlock contexts in the queue
> context cache in case of fault detection
> 
> ----------------------------------------------------------------------
> On Fri, Feb 24, 2023 at 08:39:20AM +0000, Sai Krishna Gajula wrote:
> > Hi Simon,
> >
> > > -----Original Message-----
> > > From: Simon Horman <simon.horman@corigine.com>
> > > Sent: Thursday, February 23, 2023 6:47 PM
> > > To: Sai Krishna Gajula <saikrishnag@marvell.com>
> > > Cc: davem@davemloft.net; edumazet@google.com; kuba@kernel.org;
> > > pabeni@redhat.com; netdev@vger.kernel.org;
> > > linux-kernel@vger.kernel.org; Sunil Kovvuri Goutham
> > > <sgoutham@marvell.com>; Suman Ghosh <sumang@marvell.com>
> > > Subject: Re: [net PATCH v2] octeontx2-af: Unlock contexts in the
> > > queue context cache in case of fault detection
> > >
> > >
> > > --------------------------------------------------------------------
> > > -- On Thu, Feb 23, 2023 at 04:31:25PM +0530, Sai Krishna wrote:
> > > > From: Suman Ghosh <sumang@marvell.com>
> > > >
> > > > NDC caches contexts of frequently used queue's (Rx and Tx queues)
> > > > contexts. Due to a HW errata when NDC detects fault/poision while
> > > > accessing contexts it could go into an illegal state where a cache
> > > > line could get locked forever. To makesure all cache lines in NDC
> > > > are available for optimum performance upon fault/lockerror/posion
> > > > errors scan through all cache lines in NDC and clear the lock bit.
> > > >
> > > > Fixes: 4a3581cd5995 ("octeontx2-af: NPA AQ instruction enqueue
> > > > support")
> > > > Signed-off-by: Suman Ghosh <sumang@marvell.com>
> > > > Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
> > > > Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
> > >
> > > ...
> > >
> > > > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > > b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > > index 389663a13d1d..6508f25b2b37 100644
> > > > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> > > > @@ -884,6 +884,12 @@ int rvu_cpt_lf_teardown(struct rvu *rvu, u16
> > > > pcifunc, int blkaddr, int lf,  int rvu_cpt_ctx_flush(struct rvu
> > > > *rvu,
> > > > u16 pcifunc);  int rvu_cpt_init(struct rvu *rvu);
> > > >
> > > > +/* NDC APIs */
> > > > +#define NDC_MAX_BANK(rvu, blk_addr) (rvu_read64(rvu, \
> > > > +					blk_addr, NDC_AF_CONST) & 0xFF)
> #define
> > > > +NDC_MAX_LINE_PER_BANK(rvu, blk_addr) ((rvu_read64(rvu, \
> > > > +					blk_addr, NDC_AF_CONST) &
> > > 0xFFFF0000) >> 16)
> > >
> > > Perhaps not appropriate to include as part of a fix, as NDC_MAX_BANK
> > > is being moved from elsewhere, but I wonder if this might be more
> > > cleanly implemented using FIELD_GET().
> >
> > We will modify and send a separate patch for all the possible macros that
> can be replaced by FIELD_GET().
> 
> Thanks, much appreciated.
> 
> > > ...
> > >
> > > > diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > > b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > > index 1729b22580ce..bc6ca5ccc1ff 100644
> > > > --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > > +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
> > > > @@ -694,6 +694,7 @@
> > > >  #define NDC_AF_INTR_ENA_W1S		(0x00068)
> > > >  #define NDC_AF_INTR_ENA_W1C		(0x00070)
> > > >  #define NDC_AF_ACTIVE_PC		(0x00078)
> > > > +#define NDC_AF_CAMS_RD_INTERVAL		(0x00080)
> > > >  #define NDC_AF_BP_TEST_ENABLE		(0x001F8)
> > > >  #define NDC_AF_BP_TEST(a)		(0x00200 | (a) << 3)
> > > >  #define NDC_AF_BLK_RST			(0x002F0)
> > > > @@ -709,6 +710,8 @@
> > > >  		(0x00F00 | (a) << 5 | (b) << 4)
> > > >  #define NDC_AF_BANKX_HIT_PC(a)		(0x01000 | (a) << 3)
> > > >  #define NDC_AF_BANKX_MISS_PC(a)		(0x01100 | (a) << 3)
> > > > +#define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> > > > +		(0x10000 | (a) << 3 | (b) << 3)
> > >
> > > It looks a little odd that both a and b are shifted by 3 bits.
> > > If it's intended then perhaps it would be clearer to write this as:
> > >
> > > #define NDC_AF_BANKX_LINEX_METADATA(a, b) \
> > > 		(0x10000 | ((a) | (b)) << 3)
> >
> > will send v3 patch.
> 
> Likewise, thanks.

We found a bug related to the macro NDC_AF_BANKX_LINEX_METADATA which we will fix and send in v3 patch.

Thanks,
Sai
diff mbox series

Patch

diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
index 389663a13d1d..6508f25b2b37 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
@@ -884,6 +884,12 @@  int rvu_cpt_lf_teardown(struct rvu *rvu, u16 pcifunc, int blkaddr, int lf,
 int rvu_cpt_ctx_flush(struct rvu *rvu, u16 pcifunc);
 int rvu_cpt_init(struct rvu *rvu);
 
+/* NDC APIs */
+#define NDC_MAX_BANK(rvu, blk_addr) (rvu_read64(rvu, \
+					blk_addr, NDC_AF_CONST) & 0xFF)
+#define NDC_MAX_LINE_PER_BANK(rvu, blk_addr) ((rvu_read64(rvu, \
+					blk_addr, NDC_AF_CONST) & 0xFFFF0000) >> 16)
+
 /* CN10K RVU */
 int rvu_set_channels_base(struct rvu *rvu);
 void rvu_program_channels(struct rvu *rvu);
@@ -902,6 +908,8 @@  static inline void rvu_dbg_init(struct rvu *rvu) {}
 static inline void rvu_dbg_exit(struct rvu *rvu) {}
 #endif
 
+int rvu_ndc_fix_locked_cacheline(struct rvu *rvu, int blkaddr);
+
 /* RVU Switch */
 void rvu_switch_enable(struct rvu *rvu);
 void rvu_switch_disable(struct rvu *rvu);
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c
index fa280ebd3052..fad83d1f84b0 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c
@@ -198,9 +198,6 @@  enum cpt_eng_type {
 	CPT_IE_TYPE = 3,
 };
 
-#define NDC_MAX_BANK(rvu, blk_addr) (rvu_read64(rvu, \
-						blk_addr, NDC_AF_CONST) & 0xFF)
-
 #define rvu_dbg_NULL NULL
 #define rvu_dbg_open_NULL NULL
 
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
index 26e639e57dae..4ad707e758b9 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
@@ -790,6 +790,7 @@  static int nix_aq_enqueue_wait(struct rvu *rvu, struct rvu_block *block,
 	struct nix_aq_res_s *result;
 	int timeout = 1000;
 	u64 reg, head;
+	int ret;
 
 	result = (struct nix_aq_res_s *)aq->res->base;
 
@@ -813,9 +814,22 @@  static int nix_aq_enqueue_wait(struct rvu *rvu, struct rvu_block *block,
 			return -EBUSY;
 	}
 
-	if (result->compcode != NIX_AQ_COMP_GOOD)
+	if (result->compcode != NIX_AQ_COMP_GOOD) {
 		/* TODO: Replace this with some error code */
+		if (result->compcode == NIX_AQ_COMP_CTX_FAULT ||
+		    result->compcode == NIX_AQ_COMP_LOCKERR ||
+		    result->compcode == NIX_AQ_COMP_CTX_POISON) {
+			ret = rvu_ndc_fix_locked_cacheline(rvu, BLKADDR_NDC_NIX0_RX);
+			ret |= rvu_ndc_fix_locked_cacheline(rvu, BLKADDR_NDC_NIX0_TX);
+			ret |= rvu_ndc_fix_locked_cacheline(rvu, BLKADDR_NDC_NIX1_RX);
+			ret |= rvu_ndc_fix_locked_cacheline(rvu, BLKADDR_NDC_NIX1_TX);
+			if (ret)
+				dev_err(rvu->dev,
+					"%s: Not able to unlock cachelines\n", __func__);
+		}
+
 		return -EBUSY;
+	}
 
 	return 0;
 }
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_npa.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_npa.c
index 70bd036ed76e..6cd8cc8f3488 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_npa.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_npa.c
@@ -42,9 +42,18 @@  static int npa_aq_enqueue_wait(struct rvu *rvu, struct rvu_block *block,
 			return -EBUSY;
 	}
 
-	if (result->compcode != NPA_AQ_COMP_GOOD)
+	if (result->compcode != NPA_AQ_COMP_GOOD) {
 		/* TODO: Replace this with some error code */
+		if (result->compcode == NPA_AQ_COMP_CTX_FAULT ||
+		    result->compcode == NPA_AQ_COMP_LOCKERR ||
+		    result->compcode == NPA_AQ_COMP_CTX_POISON) {
+			if (rvu_ndc_fix_locked_cacheline(rvu, BLKADDR_NDC_NPA0))
+				dev_err(rvu->dev,
+					"%s: Not able to unlock cachelines\n", __func__);
+		}
+
 		return -EBUSY;
+	}
 
 	return 0;
 }
@@ -545,3 +554,47 @@  void rvu_npa_lf_teardown(struct rvu *rvu, u16 pcifunc, int npalf)
 
 	npa_ctx_free(rvu, pfvf);
 }
+
+/* Due to an Hardware errata, in some corner cases, AQ context lock
+ * operations can result in a NDC way getting into an illegal state
+ * of not valid but locked.
+ *
+ * This API solves the problem by clearing the lock bit of the NDC block.
+ * The operation needs to be done for each line of all the NDC banks.
+ */
+int rvu_ndc_fix_locked_cacheline(struct rvu *rvu, int blkaddr)
+{
+	int bank, max_bank, line, max_line, err;
+	u64 reg;
+
+	/* Set the ENABLE bit(63) to '0' */
+	reg = rvu_read64(rvu, blkaddr, NDC_AF_CAMS_RD_INTERVAL);
+	rvu_write64(rvu, blkaddr, NDC_AF_CAMS_RD_INTERVAL, reg & GENMASK_ULL(62, 0));
+
+	/* Poll until the BUSY bits(47:32) are set to '0' */
+	err = rvu_poll_reg(rvu, blkaddr, NDC_AF_CAMS_RD_INTERVAL, GENMASK_ULL(47, 32), true);
+	if (err) {
+		dev_err(rvu->dev, "Timed out while polling for NDC CAM busy bits.\n");
+		return err;
+	}
+
+	max_bank = NDC_MAX_BANK(rvu, blkaddr);
+	max_line = NDC_MAX_LINE_PER_BANK(rvu, blkaddr);
+	for (bank = 0; bank < max_bank; bank++) {
+		for (line = 0; line < max_line; line++) {
+			/* Check if 'cache line valid bit(63)' is not set
+			 * but 'cache line lock bit(60)' is set and on
+			 * success, reset the lock bit(60).
+			 */
+			reg = rvu_read64(rvu, blkaddr,
+					 NDC_AF_BANKX_LINEX_METADATA(bank, line));
+			if (!(reg & BIT_ULL(63)) && (reg & BIT_ULL(60))) {
+				rvu_write64(rvu, blkaddr,
+					    NDC_AF_BANKX_LINEX_METADATA(bank, line),
+					    reg & ~BIT_ULL(60));
+			}
+		}
+	}
+
+	return 0;
+}
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
index 1729b22580ce..bc6ca5ccc1ff 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
@@ -694,6 +694,7 @@ 
 #define NDC_AF_INTR_ENA_W1S		(0x00068)
 #define NDC_AF_INTR_ENA_W1C		(0x00070)
 #define NDC_AF_ACTIVE_PC		(0x00078)
+#define NDC_AF_CAMS_RD_INTERVAL		(0x00080)
 #define NDC_AF_BP_TEST_ENABLE		(0x001F8)
 #define NDC_AF_BP_TEST(a)		(0x00200 | (a) << 3)
 #define NDC_AF_BLK_RST			(0x002F0)
@@ -709,6 +710,8 @@ 
 		(0x00F00 | (a) << 5 | (b) << 4)
 #define NDC_AF_BANKX_HIT_PC(a)		(0x01000 | (a) << 3)
 #define NDC_AF_BANKX_MISS_PC(a)		(0x01100 | (a) << 3)
+#define NDC_AF_BANKX_LINEX_METADATA(a, b) \
+		(0x10000 | (a) << 3 | (b) << 3)
 
 /* LBK */
 #define LBK_CONST			(0x10ull)