[PATCH net-next 00/12] nfp: bpf: support direct packet access

[PATCH net-next 00/12] nfp: bpf: support direct packet access

[Jakub Kicinski]

Hi!

The core of this series is direct packet access support.  With a
small change to the verifier, the offloaded code can now make
use of DPA.  We need to be careful to use kernel (after initial
translation) offsets in our JIT.  Direct packet access also brings
us to the problem of eBPF endianness.  After considering the 
changes necessary we decided to not support translation on both
BE and LE hosts, for now.

This series contains two fixes - one for compare instructions and
one for ineffective jne optimization.  I chose to include fixes
in this set because the code in -net works only with unreleased
PoC FW (ABI version 1) and therefore nobody outside of Netronome
can exercise it anyway.


Jakub Kicinski (12):
  bpf: verifier: set reg_type on context accesses in second pass
  nfp: bpf: reorder arguments to emit_ld_field_any()
  nfp: bpf: add missing return in jne_imm optimization
  nfp: bpf: fix compare instructions
  nfp: bpf: add mov helper
  nfp: bpf: implement byte swap instruction
  nfp: bpf: support BPF offload only on little endian
  nfp: bpf: fix context accesses
  nfp: bpf: separate I/O from checks for legacy data load
  nfp: bpf: add support for direct packet access - read
  nfp: bpf: direct packet access - write
  nfp: bpf: support direct packet access in TC

 drivers/net/ethernet/netronome/nfp/bpf/jit.c      | 402 ++++++++++++++++++----
 drivers/net/ethernet/netronome/nfp/bpf/main.c     |   2 +
 drivers/net/ethernet/netronome/nfp/bpf/main.h     |   3 +
 drivers/net/ethernet/netronome/nfp/bpf/verifier.c |  21 +-
 drivers/net/ethernet/netronome/nfp/nfp_asm.c      |   5 +-
 drivers/net/ethernet/netronome/nfp/nfp_asm.h      |   6 +-
 kernel/bpf/verifier.c                             |  43 ++-
 7 files changed, 393 insertions(+), 89 deletions(-)

-- 
2.14.1

[PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass

[Jakub Kicinski]

Use a simplified is_valid_access() callback when verifier
is used for program analysis by non-host JITs.  This allows
us to teach the verifier about packet start and packet end
offsets for direct packet access.

We can extend the callback as needed but for most packet
processing needs there isn't much more the offloads may
require.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
CC: alexei.starovoitov@gmail.com
CC: daniel@iogearbox.net

 kernel/bpf/verifier.c | 43 +++++++++++++++++++++++++++++++++++++------
 1 file changed, 37 insertions(+), 6 deletions(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 2cdbcc4f8f6b..9755279d94cb 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -813,6 +813,36 @@ static int check_packet_access(struct bpf_verifier_env *env, u32 regno, int off,
 	return err;
 }
 
+static bool analyzer_is_valid_access(struct bpf_verifier_env *env, int off,
+				     struct bpf_insn_access_aux *info)
+{
+	switch (env->prog->type) {
+	case BPF_PROG_TYPE_XDP:
+		switch (off) {
+		case offsetof(struct xdp_buff, data):
+			info->reg_type = PTR_TO_PACKET;
+			return true;
+		case offsetof(struct xdp_buff, data_end):
+			info->reg_type = PTR_TO_PACKET_END;
+			return true;
+		}
+		return false;
+	case BPF_PROG_TYPE_SCHED_CLS:
+		switch (off) {
+		case offsetof(struct sk_buff, data):
+			info->reg_type = PTR_TO_PACKET;
+			return true;
+		case offsetof(struct sk_buff, cb) +
+		     offsetof(struct bpf_skb_data_end, data_end):
+			info->reg_type = PTR_TO_PACKET_END;
+			return true;
+		}
+		return false;
+	default:
+		return false;
+	}
+}
+
 /* check access to 'struct bpf_context' fields.  Supports fixed offsets only */
 static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off, int size,
 			    enum bpf_access_type t, enum bpf_reg_type *reg_type)
@@ -821,12 +851,13 @@ static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off,
 		.reg_type = *reg_type,
 	};
 
-	/* for analyzer ctx accesses are already validated and converted */
-	if (env->analyzer_ops)
-		return 0;
-
-	if (env->prog->aux->ops->is_valid_access &&
-	    env->prog->aux->ops->is_valid_access(off, size, t, &info)) {
+	if (env->analyzer_ops) {
+		if (analyzer_is_valid_access(env, off, &info)) {
+			*reg_type = info.reg_type;
+			return 0;
+		}
+	} else if (env->prog->aux->ops->is_valid_access &&
+		   env->prog->aux->ops->is_valid_access(off, size, t, &info)) {
 		/* A non zero info.ctx_field_size indicates that this field is a
 		 * candidate for later verifier transformation to load the whole
 		 * field and then apply a mask when accessed with a narrower
-- 
2.14.1

Re: [PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass

[Daniel Borkmann]

On 10/12/2017 07:34 PM, Jakub Kicinski wrote:
> Use a simplified is_valid_access() callback when verifier
> is used for program analysis by non-host JITs.  This allows
> us to teach the verifier about packet start and packet end
> offsets for direct packet access.
>
> We can extend the callback as needed but for most packet
> processing needs there isn't much more the offloads may
> require.
>
> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> Reviewed-by: Simon Horman <simon.horman@netronome.com>
> ---
> CC: alexei.starovoitov@gmail.com
> CC: daniel@iogearbox.net
>
>   kernel/bpf/verifier.c | 43 +++++++++++++++++++++++++++++++++++++------
>   1 file changed, 37 insertions(+), 6 deletions(-)
>
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 2cdbcc4f8f6b..9755279d94cb 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -813,6 +813,36 @@ static int check_packet_access(struct bpf_verifier_env *env, u32 regno, int off,
>   	return err;
>   }
>
> +static bool analyzer_is_valid_access(struct bpf_verifier_env *env, int off,
> +				     struct bpf_insn_access_aux *info)
> +{
> +	switch (env->prog->type) {
> +	case BPF_PROG_TYPE_XDP:
> +		switch (off) {
> +		case offsetof(struct xdp_buff, data):
> +			info->reg_type = PTR_TO_PACKET;
> +			return true;
> +		case offsetof(struct xdp_buff, data_end):
> +			info->reg_type = PTR_TO_PACKET_END;
> +			return true;
> +		}
> +		return false;
> +	case BPF_PROG_TYPE_SCHED_CLS:
> +		switch (off) {
> +		case offsetof(struct sk_buff, data):
> +			info->reg_type = PTR_TO_PACKET;
> +			return true;
> +		case offsetof(struct sk_buff, cb) +
> +		     offsetof(struct bpf_skb_data_end, data_end):
> +			info->reg_type = PTR_TO_PACKET_END;
> +			return true;
> +		}
> +		return false;
> +	default:
> +		return false;
> +	}
> +}
> +
>   /* check access to 'struct bpf_context' fields.  Supports fixed offsets only */
>   static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off, int size,
>   			    enum bpf_access_type t, enum bpf_reg_type *reg_type)
> @@ -821,12 +851,13 @@ static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off,
>   		.reg_type = *reg_type,
>   	};
>
> -	/* for analyzer ctx accesses are already validated and converted */
> -	if (env->analyzer_ops)
> -		return 0;
> -
> -	if (env->prog->aux->ops->is_valid_access &&
> -	    env->prog->aux->ops->is_valid_access(off, size, t, &info)) {
> +	if (env->analyzer_ops) {
> +		if (analyzer_is_valid_access(env, off, &info)) {
> +			*reg_type = info.reg_type;

Is there some specific issue with the is_valid_access() callbacks that you
need to do this (I couldn't parse that out of the commit message)? It would
be nice to keep the reg_type setting in one place, meaning the callbacks
themselves, so we wouldn't need to maintain this in multiple places.

> +			return 0;
> +		}
> +	} else if (env->prog->aux->ops->is_valid_access &&
> +		   env->prog->aux->ops->is_valid_access(off, size, t, &info)) {
>   		/* A non zero info.ctx_field_size indicates that this field is a
>   		 * candidate for later verifier transformation to load the whole
>   		 * field and then apply a mask when accessed with a narrower
>

Re: [PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass

[Jakub Kicinski]

On Thu, 12 Oct 2017 22:43:10 +0200, Daniel Borkmann wrote:
> On 10/12/2017 07:34 PM, Jakub Kicinski wrote:
> > Use a simplified is_valid_access() callback when verifier
> > is used for program analysis by non-host JITs.  This allows
> > us to teach the verifier about packet start and packet end
> > offsets for direct packet access.
> >
> > We can extend the callback as needed but for most packet
> > processing needs there isn't much more the offloads may
> > require.
> >
> > Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> > Reviewed-by: Simon Horman <simon.horman@netronome.com>
> > ---
> > CC: alexei.starovoitov@gmail.com
> > CC: daniel@iogearbox.net
> >
> >   kernel/bpf/verifier.c | 43 +++++++++++++++++++++++++++++++++++++------
> >   1 file changed, 37 insertions(+), 6 deletions(-)
> >
> > diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> > index 2cdbcc4f8f6b..9755279d94cb 100644
> > --- a/kernel/bpf/verifier.c
> > +++ b/kernel/bpf/verifier.c
> > @@ -813,6 +813,36 @@ static int check_packet_access(struct bpf_verifier_env *env, u32 regno, int off,
> >   	return err;
> >   }
> >
> > +static bool analyzer_is_valid_access(struct bpf_verifier_env *env, int off,
> > +				     struct bpf_insn_access_aux *info)
> > +{
> > +	switch (env->prog->type) {
> > +	case BPF_PROG_TYPE_XDP:
> > +		switch (off) {
> > +		case offsetof(struct xdp_buff, data):
> > +			info->reg_type = PTR_TO_PACKET;
> > +			return true;
> > +		case offsetof(struct xdp_buff, data_end):
> > +			info->reg_type = PTR_TO_PACKET_END;
> > +			return true;
> > +		}
> > +		return false;
> > +	case BPF_PROG_TYPE_SCHED_CLS:
> > +		switch (off) {
> > +		case offsetof(struct sk_buff, data):
> > +			info->reg_type = PTR_TO_PACKET;
> > +			return true;
> > +		case offsetof(struct sk_buff, cb) +
> > +		     offsetof(struct bpf_skb_data_end, data_end):
> > +			info->reg_type = PTR_TO_PACKET_END;
> > +			return true;
> > +		}
> > +		return false;
> > +	default:
> > +		return false;
> > +	}
> > +}
> > +
> >   /* check access to 'struct bpf_context' fields.  Supports fixed offsets only */
> >   static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off, int size,
> >   			    enum bpf_access_type t, enum bpf_reg_type *reg_type)
> > @@ -821,12 +851,13 @@ static int check_ctx_access(struct bpf_verifier_env *env, int insn_idx, int off,
> >   		.reg_type = *reg_type,
> >   	};
> >
> > -	/* for analyzer ctx accesses are already validated and converted */
> > -	if (env->analyzer_ops)
> > -		return 0;
> > -
> > -	if (env->prog->aux->ops->is_valid_access &&
> > -	    env->prog->aux->ops->is_valid_access(off, size, t, &info)) {
> > +	if (env->analyzer_ops) {
> > +		if (analyzer_is_valid_access(env, off, &info)) {
> > +			*reg_type = info.reg_type;  
> 
> Is there some specific issue with the is_valid_access() callbacks that you
> need to do this (I couldn't parse that out of the commit message)?

Do you mean why not just call is_valid_access()?  The offsets are
translated, so is_valid_access() will use user space __sk_buff's
offsets while we have the kernel's sk_buff offsets here...

> It would be nice to keep the reg_type setting in one place, meaning
> the callbacks themselves, so we wouldn't need to maintain this in
> multiple places.

Hm.. I though this was the smallest and simplest change.  I could
translate the offsets but that seems wobbly.  Or try to consolidate the
call into the same if () branch?  Not sure..

As a bonus info I discovered there is a bug in -net with how things are
converted.  We allow arithmetic on context pointers but then only
look at the insn.off in the converter...  I'm working on a fix.

Re: [PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass

[Daniel Borkmann]

On 10/12/2017 10:56 PM, Jakub Kicinski wrote:
> On Thu, 12 Oct 2017 22:43:10 +0200, Daniel Borkmann wrote:
[...]
>> It would be nice to keep the reg_type setting in one place, meaning
>> the callbacks themselves, so we wouldn't need to maintain this in
>> multiple places.
>
> Hm.. I though this was the smallest and simplest change.  I could
> translate the offsets but that seems wobbly.  Or try to consolidate the
> call into the same if () branch?  Not sure..

Different callbacks for post-verification would be good at min as it
would allow to keep all the context access info in one place for a
given type at least.

> As a bonus info I discovered there is a bug in -net with how things are
> converted.  We allow arithmetic on context pointers but then only
> look at the insn.off in the converter...  I'm working on a fix.

Ohh well, good catch, indeed! :( Can you also add coverage to the
bpf selftests for this?

Thanks,
Daniel

Re: [PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass

[Jakub Kicinski]

On Thu, 12 Oct 2017 23:33:21 +0200, Daniel Borkmann wrote:
> On 10/12/2017 10:56 PM, Jakub Kicinski wrote:
> > On Thu, 12 Oct 2017 22:43:10 +0200, Daniel Borkmann wrote:  
> [...]
> >> It would be nice to keep the reg_type setting in one place, meaning
> >> the callbacks themselves, so we wouldn't need to maintain this in
> >> multiple places.  
> >
> > Hm.. I though this was the smallest and simplest change.  I could
> > translate the offsets but that seems wobbly.  Or try to consolidate the
> > call into the same if () branch?  Not sure..  
> 
> Different callbacks for post-verification would be good at min as it
> would allow to keep all the context access info in one place for a
> given type at least.

Sorry to be clear - you're suggesting adding a new callback to struct
bpf_verifier_ops, or swapping the struct bpf_verifier_ops for a
special post-verification one?

> > As a bonus info I discovered there is a bug in -net with how things are
> > converted.  We allow arithmetic on context pointers but then only
> > look at the insn.off in the converter...  I'm working on a fix.  
> 
> Ohh well, good catch, indeed! :( Can you also add coverage to the
> bpf selftests for this?

Will do!

Re: [PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass

[Daniel Borkmann]

On 10/12/2017 11:39 PM, Jakub Kicinski wrote:
> On Thu, 12 Oct 2017 23:33:21 +0200, Daniel Borkmann wrote:
>> On 10/12/2017 10:56 PM, Jakub Kicinski wrote:
>>> On Thu, 12 Oct 2017 22:43:10 +0200, Daniel Borkmann wrote:
>> [...]
>>>> It would be nice to keep the reg_type setting in one place, meaning
>>>> the callbacks themselves, so we wouldn't need to maintain this in
>>>> multiple places.
>>>
>>> Hm.. I though this was the smallest and simplest change.  I could
>>> translate the offsets but that seems wobbly.  Or try to consolidate the
>>> call into the same if () branch?  Not sure..
>>
>> Different callbacks for post-verification would be good at min as it
>> would allow to keep all the context access info in one place for a
>> given type at least.
>
> Sorry to be clear - you're suggesting adding a new callback to struct
> bpf_verifier_ops, or swapping the struct bpf_verifier_ops for a
> special post-verification one?

Either way is fine by me.

>>> As a bonus info I discovered there is a bug in -net with how things are
>>> converted.  We allow arithmetic on context pointers but then only
>>> look at the insn.off in the converter...  I'm working on a fix.
>>
>> Ohh well, good catch, indeed! :( Can you also add coverage to the
>> bpf selftests for this?
>
> Will do!

Thanks,
Daniel

[PATCH net-next 02/12] nfp: bpf: reorder arguments to emit_ld_field_any()

[Jakub Kicinski]

ld_field instruction has the following format in NFP assembler:

  ld_field[dst, 1000, src, <<24]

reoder parameters to emit_ld_field_any() to make it closer to
the familiar assembler order.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 13148f30fc4c..cf8a6eb3ec99 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -402,8 +402,8 @@ __emit_ld_field(struct nfp_prog *nfp_prog, enum shf_sc sc,
 }
 
 static void
-emit_ld_field_any(struct nfp_prog *nfp_prog, enum shf_sc sc, u8 shift,
-		  swreg dst, u8 bmask, swreg src, bool zero)
+emit_ld_field_any(struct nfp_prog *nfp_prog, swreg dst, u8 bmask, swreg src,
+		  enum shf_sc sc, u8 shift, bool zero)
 {
 	struct nfp_insn_re_regs reg;
 	int err;
@@ -424,7 +424,7 @@ static void
 emit_ld_field(struct nfp_prog *nfp_prog, swreg dst, u8 bmask, swreg src,
 	      enum shf_sc sc, u8 shift)
 {
-	emit_ld_field_any(nfp_prog, sc, shift, dst, bmask, src, false);
+	emit_ld_field_any(nfp_prog, dst, bmask, src, sc, shift, false);
 }
 
 static void emit_nop(struct nfp_prog *nfp_prog)
-- 
2.14.1

[PATCH net-next 03/12] nfp: bpf: add missing return in jne_imm optimization

[Jakub Kicinski]

We optimize comparisons to immediate 0 as if (reg.lo | reg.hi).
The early return statement was missing, however, which means we
would generate two comparisons - optimized one followed by a
normal 2x 32 bit compare.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index cf8a6eb3ec99..5ac834e91aed 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -1191,6 +1191,7 @@ static int jne_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 		emit_alu(nfp_prog, reg_none(), reg_a(insn->dst_reg * 2),
 			 ALU_OP_OR, reg_b(insn->dst_reg * 2 + 1));
 		emit_br(nfp_prog, BR_BNE, insn->off, 0);
+		return 0;
 	}
 
 	tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));
-- 
2.14.1

[PATCH net-next 04/12] nfp: bpf: fix compare instructions

[Jakub Kicinski]

Now that we have BPF assemebler support in LLVM 6 we can easily
test all compare instructions (LLVM 4 didn't generate most of them
from C).  Fix the compare to immediates and refactor the order
of compare to regs to make sure they both follow the same pattern.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 5ac834e91aed..e970f284c8a4 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -720,7 +720,10 @@ wrp_cmp_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
 	    enum br_mask br_mask, bool swap)
 {
 	const struct bpf_insn *insn = &meta->insn;
-	u8 areg = insn->src_reg * 2, breg = insn->dst_reg * 2;
+	u8 areg, breg;
+
+	areg = insn->dst_reg * 2;
+	breg = insn->src_reg * 2;
 
 	if (insn->off < 0) /* TODO */
 		return -EOPNOTSUPP;
@@ -1129,22 +1132,22 @@ static int jeq_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 
 static int jgt_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_imm(nfp_prog, meta, BR_BLO, false);
+	return wrp_cmp_imm(nfp_prog, meta, BR_BLO, true);
 }
 
 static int jge_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_imm(nfp_prog, meta, BR_BHS, true);
+	return wrp_cmp_imm(nfp_prog, meta, BR_BHS, false);
 }
 
 static int jlt_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_imm(nfp_prog, meta, BR_BHS, false);
+	return wrp_cmp_imm(nfp_prog, meta, BR_BLO, false);
 }
 
 static int jle_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_imm(nfp_prog, meta, BR_BLO, true);
+	return wrp_cmp_imm(nfp_prog, meta, BR_BHS, true);
 }
 
 static int jset_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
@@ -1227,22 +1230,22 @@ static int jeq_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 
 static int jgt_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_reg(nfp_prog, meta, BR_BLO, false);
+	return wrp_cmp_reg(nfp_prog, meta, BR_BLO, true);
 }
 
 static int jge_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_reg(nfp_prog, meta, BR_BHS, true);
+	return wrp_cmp_reg(nfp_prog, meta, BR_BHS, false);
 }
 
 static int jlt_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_reg(nfp_prog, meta, BR_BHS, false);
+	return wrp_cmp_reg(nfp_prog, meta, BR_BLO, false);
 }
 
 static int jle_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	return wrp_cmp_reg(nfp_prog, meta, BR_BLO, true);
+	return wrp_cmp_reg(nfp_prog, meta, BR_BHS, true);
 }
 
 static int jset_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
-- 
2.14.1

[PATCH net-next 05/12] nfp: bpf: add mov helper

[Jakub Kicinski]

Register move operation is encoded as alu no op.  This means
that one has to specify number of unused/none parameters to
the emit_alu().  Add a helper.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 31 ++++++++++++++--------------
 1 file changed, 15 insertions(+), 16 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index e970f284c8a4..4f7cfa6adfc1 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -504,9 +504,14 @@ wrp_br_special(struct nfp_prog *nfp_prog, enum br_mask mask,
 		FIELD_PREP(OP_BR_SPECIAL, special);
 }
 
+static void wrp_mov(struct nfp_prog *nfp_prog, swreg dst, swreg src)
+{
+	emit_alu(nfp_prog, dst, reg_none(), ALU_OP_NONE, src);
+}
+
 static void wrp_reg_mov(struct nfp_prog *nfp_prog, u16 dst, u16 src)
 {
-	emit_alu(nfp_prog, reg_both(dst), reg_none(), ALU_OP_NONE, reg_b(src));
+	wrp_mov(nfp_prog, reg_both(dst), reg_b(src));
 }
 
 static int
@@ -556,8 +561,7 @@ construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset,
 			 reg_xfer(0), SHF_SC_R_SHF, shift * 8);
 	else
 		for (; i * 4 < size; i++)
-			emit_alu(nfp_prog, reg_both(i),
-				 reg_none(), ALU_OP_NONE, reg_xfer(i));
+			wrp_mov(nfp_prog, reg_both(i), reg_xfer(i));
 
 	if (i < 2)
 		wrp_immed(nfp_prog, reg_both(1), 0);
@@ -1032,8 +1036,8 @@ static int data_ind_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 static int mem_ldx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
 	if (meta->insn.off == offsetof(struct sk_buff, len))
-		emit_alu(nfp_prog, reg_both(meta->insn.dst_reg * 2),
-			 reg_none(), ALU_OP_NONE, plen_reg(nfp_prog));
+		wrp_mov(nfp_prog,
+			reg_both(meta->insn.dst_reg * 2), plen_reg(nfp_prog));
 	else
 		return -EOPNOTSUPP;
 
@@ -1048,7 +1052,7 @@ static int mem_ldx4_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 	    meta->insn.off != offsetof(struct xdp_md, data_end))
 		return -EOPNOTSUPP;
 
-	emit_alu(nfp_prog, dst, reg_none(), ALU_OP_NONE, pptr_reg(nfp_prog));
+	wrp_mov(nfp_prog, dst, pptr_reg(nfp_prog));
 
 	if (meta->insn.off == offsetof(struct xdp_md, data))
 		return 0;
@@ -1438,8 +1442,7 @@ static void nfp_outro_tc_legacy(struct nfp_prog *nfp_prog)
 	 *  ife + tx  0x24 -> redir, count as stat1
 	 */
 	emit_br_byte_neq(nfp_prog, reg_b(0), 0xff, 0, nfp_prog->tgt_done, 2);
-	emit_alu(nfp_prog, reg_a(0),
-		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_FLAGS);
+	wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS);
 	emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(0x11), SHF_SC_L_SHF, 16);
 
 	emit_br(nfp_prog, BR_UNC, nfp_prog->tgt_done, 1);
@@ -1466,8 +1469,7 @@ static void nfp_outro_tc_da(struct nfp_prog *nfp_prog)
 
 	emit_br_def(nfp_prog, nfp_prog->tgt_done, 2);
 
-	emit_alu(nfp_prog, reg_a(0),
-		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_FLAGS);
+	wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS);
 	emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(0x11), SHF_SC_L_SHF, 16);
 
 	/* Target for normal exits */
@@ -1476,8 +1478,7 @@ static void nfp_outro_tc_da(struct nfp_prog *nfp_prog)
 	/* if R0 > 7 jump to abort */
 	emit_alu(nfp_prog, reg_none(), reg_imm(7), ALU_OP_SUB, reg_b(0));
 	emit_br(nfp_prog, BR_BLO, nfp_prog->tgt_abort, 0);
-	emit_alu(nfp_prog, reg_a(0),
-		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_FLAGS);
+	wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS);
 
 	wrp_immed(nfp_prog, reg_b(2), 0x41221211);
 	wrp_immed(nfp_prog, reg_b(3), 0x41001211);
@@ -1514,8 +1515,7 @@ static void nfp_outro_xdp(struct nfp_prog *nfp_prog)
 
 	emit_br_def(nfp_prog, nfp_prog->tgt_done, 2);
 
-	emit_alu(nfp_prog, reg_a(0),
-		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_FLAGS);
+	wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS);
 	emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(0x82), SHF_SC_L_SHF, 16);
 
 	/* Target for normal exits */
@@ -1536,8 +1536,7 @@ static void nfp_outro_xdp(struct nfp_prog *nfp_prog)
 
 	emit_br_def(nfp_prog, nfp_prog->tgt_done, 2);
 
-	emit_alu(nfp_prog, reg_a(0),
-		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_FLAGS);
+	wrp_mov(nfp_prog, reg_a(0), NFP_BPF_ABI_FLAGS);
 	emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_b(2), SHF_SC_L_SHF, 16);
 }
 
-- 
2.14.1

[PATCH net-next 06/12] nfp: bpf: implement byte swap instruction

[Jakub Kicinski]

Implement byte swaps with rotations, shifts and byte loads.
Remember to clear upper parts of the 64 bit registers.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 38 ++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 4f7cfa6adfc1..5e8a6b766790 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -746,6 +746,14 @@ wrp_cmp_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
 	return 0;
 }
 
+static void wrp_end32(struct nfp_prog *nfp_prog, swreg reg_in, u8 gpr_out)
+{
+	emit_ld_field(nfp_prog, reg_both(gpr_out), 0xf, reg_in,
+		      SHF_SC_R_ROT, 8);
+	emit_ld_field(nfp_prog, reg_both(gpr_out), 0x5, reg_a(gpr_out),
+		      SHF_SC_R_ROT, 16);
+}
+
 /* --- Callbacks --- */
 static int mov_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
@@ -982,6 +990,35 @@ static int shl_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 	return 0;
 }
 
+static int end_reg32(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u8 gpr = insn->dst_reg * 2;
+
+	switch (insn->imm) {
+	case 16:
+		emit_ld_field(nfp_prog, reg_both(gpr), 0x9, reg_b(gpr),
+			      SHF_SC_R_ROT, 8);
+		emit_ld_field(nfp_prog, reg_both(gpr), 0xe, reg_a(gpr),
+			      SHF_SC_R_SHF, 16);
+
+		wrp_immed(nfp_prog, reg_both(gpr + 1), 0);
+		break;
+	case 32:
+		wrp_end32(nfp_prog, reg_a(gpr), gpr);
+		wrp_immed(nfp_prog, reg_both(gpr + 1), 0);
+		break;
+	case 64:
+		wrp_mov(nfp_prog, imm_a(nfp_prog), reg_b(gpr + 1));
+
+		wrp_end32(nfp_prog, reg_a(gpr), gpr + 1);
+		wrp_end32(nfp_prog, imm_a(nfp_prog), gpr);
+		break;
+	}
+
+	return 0;
+}
+
 static int imm_ld8_part2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
 	wrp_immed(nfp_prog, reg_both(nfp_meta_prev(meta)->insn.dst_reg * 2 + 1),
@@ -1297,6 +1334,7 @@ static const instr_cb_t instr_cb[256] = {
 	[BPF_ALU | BPF_SUB | BPF_X] =	sub_reg,
 	[BPF_ALU | BPF_SUB | BPF_K] =	sub_imm,
 	[BPF_ALU | BPF_LSH | BPF_K] =	shl_imm,
+	[BPF_ALU | BPF_END | BPF_X] =	end_reg32,
 	[BPF_LD | BPF_IMM | BPF_DW] =	imm_ld8,
 	[BPF_LD | BPF_ABS | BPF_B] =	data_ld1,
 	[BPF_LD | BPF_ABS | BPF_H] =	data_ld2,
-- 
2.14.1

[PATCH net-next 07/12] nfp: bpf: support BPF offload only on little endian

[Jakub Kicinski]

eBPF is host-endian specific.  Translating both BE and LE eBPF
to the NFP is feasible, but would require quite a bit of indirection.
The fact that I don't have access to any BE hosts that would fit
a 25G/40G/100G NIC is also limiting my ability to test big endian.

For now restrict the offload to little endian hosts only.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/main.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.c b/drivers/net/ethernet/netronome/nfp/bpf/main.c
index 074726980994..6e74f8db1cc1 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/main.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/main.c
@@ -42,9 +42,11 @@
 
 static bool nfp_net_ebpf_capable(struct nfp_net *nn)
 {
+#ifdef __LITTLE_ENDIAN
 	if (nn->cap & NFP_NET_CFG_CTRL_BPF &&
 	    nn_readb(nn, NFP_NET_CFG_BPF_ABI) == NFP_NET_BPF_ABI)
 		return true;
+#endif
 	return false;
 }
 
-- 
2.14.1

[PATCH net-next 08/12] nfp: bpf: fix context accesses

[Jakub Kicinski]

Sizes of fields in struct xdp_md/xdp_buff and some in sk_buff depend
on target architecture.  Take that into account and use struct xdp_buff,
not struct xdp_md.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 49 ++++++++++++++++------------
 1 file changed, 29 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 5e8a6b766790..4b62f5497728 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -1070,47 +1070,56 @@ static int data_ind_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 				     meta->insn.src_reg * 2, true, 4);
 }
 
-static int mem_ldx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+static int mem_ldx_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+		       u8 size)
 {
-	if (meta->insn.off == offsetof(struct sk_buff, len))
+	switch (meta->insn.off) {
+	case offsetof(struct sk_buff, len):
+		if (size != FIELD_SIZEOF(struct sk_buff, len))
+			return -EOPNOTSUPP;
 		wrp_mov(nfp_prog,
 			reg_both(meta->insn.dst_reg * 2), plen_reg(nfp_prog));
-	else
+		break;
+	default:
 		return -EOPNOTSUPP;
+	}
+
+	wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), 0);
 
 	return 0;
 }
 
-static int mem_ldx4_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+static int mem_ldx_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+		       u8 size)
 {
 	swreg dst = reg_both(meta->insn.dst_reg * 2);
 
-	if (meta->insn.off != offsetof(struct xdp_md, data) &&
-	    meta->insn.off != offsetof(struct xdp_md, data_end))
-		return -EOPNOTSUPP;
-
-	wrp_mov(nfp_prog, dst, pptr_reg(nfp_prog));
+	if (size != sizeof(void *))
+		return -EINVAL;
 
-	if (meta->insn.off == offsetof(struct xdp_md, data))
-		return 0;
+	switch (meta->insn.off) {
+	case offsetof(struct xdp_buff, data):
+		wrp_mov(nfp_prog, dst, pptr_reg(nfp_prog));
+		break;
+	case offsetof(struct xdp_buff, data_end):
+		emit_alu(nfp_prog, dst,
+			 plen_reg(nfp_prog), ALU_OP_ADD, pptr_reg(nfp_prog));
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
 
-	emit_alu(nfp_prog, dst,	dst, ALU_OP_ADD, plen_reg(nfp_prog));
+	wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), 0);
 
 	return 0;
 }
 
 static int mem_ldx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	int ret;
-
 	if (nfp_prog->act == NN_ACT_XDP)
-		ret = mem_ldx4_xdp(nfp_prog, meta);
+		return mem_ldx_xdp(nfp_prog, meta, 4);
 	else
-		ret = mem_ldx4_skb(nfp_prog, meta);
-
-	wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), 0);
-
-	return ret;
+		return mem_ldx_skb(nfp_prog, meta, 4);
 }
 
 static int mem_stx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
-- 
2.14.1

[PATCH net-next 09/12] nfp: bpf: separate I/O from checks for legacy data load

[Jakub Kicinski]

Move data load into a separate function and separate it from
packet length checks of legacy I/O.  This makes the code more
readable and easier to reuse.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 77 +++++++++++++++-------------
 1 file changed, 40 insertions(+), 37 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 4b62f5497728..3e173da16428 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -515,63 +515,66 @@ static void wrp_reg_mov(struct nfp_prog *nfp_prog, u16 dst, u16 src)
 }
 
 static int
-construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset,
-		      u16 src, bool src_valid, u8 size)
+data_ld(struct nfp_prog *nfp_prog, swreg offset, u8 dst_gpr, int size)
 {
 	unsigned int i;
 	u16 shift, sz;
-	swreg tmp_reg;
 
 	/* We load the value from the address indicated in @offset and then
 	 * shift out the data we don't need.  Note: this is big endian!
 	 */
-	sz = size < 4 ? 4 : size;
+	sz = max(size, 4);
 	shift = size < 4 ? 4 - size : 0;
 
-	if (src_valid) {
-		/* Calculate the true offset (src_reg + imm) */
-		tmp_reg = ur_load_imm_any(nfp_prog, offset, imm_b(nfp_prog));
-		emit_alu(nfp_prog, imm_both(nfp_prog),
-			 reg_a(src), ALU_OP_ADD, tmp_reg);
-		/* Check packet length (size guaranteed to fit b/c it's u8) */
-		emit_alu(nfp_prog, imm_a(nfp_prog),
-			 imm_a(nfp_prog), ALU_OP_ADD, reg_imm(size));
-		emit_alu(nfp_prog, reg_none(),
-			 plen_reg(nfp_prog), ALU_OP_SUB, imm_a(nfp_prog));
-		wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT);
-		/* Load data */
-		emit_cmd(nfp_prog, CMD_TGT_READ8, CMD_MODE_32b, 0,
-			 pptr_reg(nfp_prog), imm_b(nfp_prog), sz - 1, true);
-	} else {
-		/* Check packet length */
-		tmp_reg = ur_load_imm_any(nfp_prog, offset + size,
-					  imm_a(nfp_prog));
-		emit_alu(nfp_prog, reg_none(),
-			 plen_reg(nfp_prog), ALU_OP_SUB, tmp_reg);
-		wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT);
-		/* Load data */
-		tmp_reg = re_load_imm_any(nfp_prog, offset, imm_b(nfp_prog));
-		emit_cmd(nfp_prog, CMD_TGT_READ8, CMD_MODE_32b, 0,
-			 pptr_reg(nfp_prog), tmp_reg, sz - 1, true);
-	}
+	emit_cmd(nfp_prog, CMD_TGT_READ8, CMD_MODE_32b, 0,
+		 pptr_reg(nfp_prog), offset, sz - 1, true);
 
 	i = 0;
 	if (shift)
-		emit_shf(nfp_prog, reg_both(0), reg_none(), SHF_OP_NONE,
+		emit_shf(nfp_prog, reg_both(dst_gpr), reg_none(), SHF_OP_NONE,
 			 reg_xfer(0), SHF_SC_R_SHF, shift * 8);
 	else
 		for (; i * 4 < size; i++)
-			wrp_mov(nfp_prog, reg_both(i), reg_xfer(i));
+			wrp_mov(nfp_prog, reg_both(dst_gpr + i), reg_xfer(i));
 
 	if (i < 2)
-		wrp_immed(nfp_prog, reg_both(1), 0);
+		wrp_immed(nfp_prog, reg_both(dst_gpr + 1), 0);
 
 	return 0;
 }
 
+static int
+construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset, u16 src, u8 size)
+{
+	swreg tmp_reg;
+
+	/* Calculate the true offset (src_reg + imm) */
+	tmp_reg = ur_load_imm_any(nfp_prog, offset, imm_b(nfp_prog));
+	emit_alu(nfp_prog, imm_both(nfp_prog), reg_a(src), ALU_OP_ADD, tmp_reg);
+
+	/* Check packet length (size guaranteed to fit b/c it's u8) */
+	emit_alu(nfp_prog, imm_a(nfp_prog),
+		 imm_a(nfp_prog), ALU_OP_ADD, reg_imm(size));
+	emit_alu(nfp_prog, reg_none(),
+		 plen_reg(nfp_prog), ALU_OP_SUB, imm_a(nfp_prog));
+	wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT);
+
+	/* Load data */
+	return data_ld(nfp_prog, imm_b(nfp_prog), 0, size);
+}
+
 static int construct_data_ld(struct nfp_prog *nfp_prog, u16 offset, u8 size)
 {
-	return construct_data_ind_ld(nfp_prog, offset, 0, false, size);
+	swreg tmp_reg;
+
+	/* Check packet length */
+	tmp_reg = ur_load_imm_any(nfp_prog, offset + size, imm_a(nfp_prog));
+	emit_alu(nfp_prog, reg_none(), plen_reg(nfp_prog), ALU_OP_SUB, tmp_reg);
+	wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT);
+
+	/* Load data */
+	tmp_reg = re_load_imm_any(nfp_prog, offset, imm_b(nfp_prog));
+	return data_ld(nfp_prog, tmp_reg, 0, size);
 }
 
 static void
@@ -1055,19 +1058,19 @@ static int data_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 static int data_ind_ld1(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
 	return construct_data_ind_ld(nfp_prog, meta->insn.imm,
-				     meta->insn.src_reg * 2, true, 1);
+				     meta->insn.src_reg * 2, 1);
 }
 
 static int data_ind_ld2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
 	return construct_data_ind_ld(nfp_prog, meta->insn.imm,
-				     meta->insn.src_reg * 2, true, 2);
+				     meta->insn.src_reg * 2, 2);
 }
 
 static int data_ind_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
 	return construct_data_ind_ld(nfp_prog, meta->insn.imm,
-				     meta->insn.src_reg * 2, true, 4);
+				     meta->insn.src_reg * 2, 4);
 }
 
 static int mem_ldx_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
-- 
2.14.1

[PATCH net-next 10/12] nfp: bpf: add support for direct packet access - read

[Jakub Kicinski]

In direct packet access bound checks are already done, we can
simply dereference the packet pointer.

Verifier/parser logic needs to record pointer type.  Note that
although verifier does protect us from CTX vs other pointer
changes we will also want to differentiate between PACKET vs
MAP_VALUE or STACK, so we can add the check already.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c      | 85 +++++++++++++++++++++--
 drivers/net/ethernet/netronome/nfp/bpf/main.h     |  3 +
 drivers/net/ethernet/netronome/nfp/bpf/verifier.c | 21 ++++--
 drivers/net/ethernet/netronome/nfp/nfp_asm.c      |  3 +
 drivers/net/ethernet/netronome/nfp/nfp_asm.h      |  4 ++
 5 files changed, 105 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 3e173da16428..975d63fbc1d5 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -543,6 +543,36 @@ data_ld(struct nfp_prog *nfp_prog, swreg offset, u8 dst_gpr, int size)
 	return 0;
 }
 
+static int
+data_ld_host_order(struct nfp_prog *nfp_prog, u8 src_gpr, swreg offset,
+		   u8 dst_gpr, int size)
+{
+	unsigned int i;
+	u8 mask, sz;
+
+	/* We load the value from the address indicated in @offset and then
+	 * mask out the data we don't need.  Note: this is little endian!
+	 */
+	sz = max(size, 4);
+	mask = size < 4 ? GENMASK(size - 1, 0) : 0;
+
+	emit_cmd(nfp_prog, CMD_TGT_READ32_SWAP, CMD_MODE_32b, 0,
+		 reg_a(src_gpr), offset, sz / 4 - 1, true);
+
+	i = 0;
+	if (mask)
+		emit_ld_field_any(nfp_prog, reg_both(dst_gpr), mask,
+				  reg_xfer(0), SHF_SC_NONE, 0, true);
+	else
+		for (; i * 4 < size; i++)
+			wrp_mov(nfp_prog, reg_both(dst_gpr + i), reg_xfer(i));
+
+	if (i < 2)
+		wrp_immed(nfp_prog, reg_both(dst_gpr + 1), 0);
+
+	return 0;
+}
+
 static int
 construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset, u16 src, u8 size)
 {
@@ -1117,12 +1147,53 @@ static int mem_ldx_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
 	return 0;
 }
 
+static int
+mem_ldx_data(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	     unsigned int size)
+{
+	swreg tmp_reg;
+
+	tmp_reg = re_load_imm_any(nfp_prog, meta->insn.off, imm_b(nfp_prog));
+
+	return data_ld_host_order(nfp_prog, meta->insn.src_reg * 2, tmp_reg,
+				  meta->insn.dst_reg * 2, size);
+}
+
+static int
+mem_ldx(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	unsigned int size)
+{
+	if (meta->ptr.type == PTR_TO_CTX) {
+		if (nfp_prog->act == NN_ACT_XDP)
+			return mem_ldx_xdp(nfp_prog, meta, size);
+		else
+			return mem_ldx_skb(nfp_prog, meta, size);
+	}
+
+	if (meta->ptr.type == PTR_TO_PACKET)
+		return mem_ldx_data(nfp_prog, meta, size);
+
+	return -EOPNOTSUPP;
+}
+
+static int mem_ldx1(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_ldx(nfp_prog, meta, 1);
+}
+
+static int mem_ldx2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_ldx(nfp_prog, meta, 2);
+}
+
 static int mem_ldx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	if (nfp_prog->act == NN_ACT_XDP)
-		return mem_ldx_xdp(nfp_prog, meta, 4);
-	else
-		return mem_ldx_skb(nfp_prog, meta, 4);
+	return mem_ldx(nfp_prog, meta, 4);
+}
+
+static int mem_ldx8(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_ldx(nfp_prog, meta, 8);
 }
 
 static int mem_stx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
@@ -1137,6 +1208,9 @@ static int mem_stx4_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 
 static int mem_stx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
+	if (meta->ptr.type == PTR_TO_PACKET)
+		return -EOPNOTSUPP;
+
 	if (nfp_prog->act == NN_ACT_XDP)
 		return mem_stx4_xdp(nfp_prog, meta);
 	return mem_stx4_skb(nfp_prog, meta);
@@ -1354,7 +1428,10 @@ static const instr_cb_t instr_cb[256] = {
 	[BPF_LD | BPF_IND | BPF_B] =	data_ind_ld1,
 	[BPF_LD | BPF_IND | BPF_H] =	data_ind_ld2,
 	[BPF_LD | BPF_IND | BPF_W] =	data_ind_ld4,
+	[BPF_LDX | BPF_MEM | BPF_B] =	mem_ldx1,
+	[BPF_LDX | BPF_MEM | BPF_H] =	mem_ldx2,
 	[BPF_LDX | BPF_MEM | BPF_W] =	mem_ldx4,
+	[BPF_LDX | BPF_MEM | BPF_DW] =	mem_ldx8,
 	[BPF_STX | BPF_MEM | BPF_W] =	mem_stx4,
 	[BPF_JMP | BPF_JA | BPF_K] =	jump,
 	[BPF_JMP | BPF_JEQ | BPF_K] =	jeq_imm,
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.h b/drivers/net/ethernet/netronome/nfp/bpf/main.h
index b7a112acbdb7..d77e88a45409 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/main.h
+++ b/drivers/net/ethernet/netronome/nfp/bpf/main.h
@@ -36,6 +36,7 @@
 
 #include <linux/bitfield.h>
 #include <linux/bpf.h>
+#include <linux/bpf_verifier.h>
 #include <linux/list.h>
 #include <linux/types.h>
 
@@ -96,6 +97,7 @@ typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *);
 /**
  * struct nfp_insn_meta - BPF instruction wrapper
  * @insn: BPF instruction
+ * @ptr: pointer type for memory operations
  * @off: index of first generated machine instruction (in nfp_prog.prog)
  * @n: eBPF instruction number
  * @skip: skip this instruction (optimized out)
@@ -104,6 +106,7 @@ typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *);
  */
 struct nfp_insn_meta {
 	struct bpf_insn insn;
+	struct bpf_reg_state ptr;
 	unsigned int off;
 	unsigned short n;
 	bool skip;
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/verifier.c b/drivers/net/ethernet/netronome/nfp/bpf/verifier.c
index 5b783a91b115..e361c0e3b788 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/verifier.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/verifier.c
@@ -112,12 +112,19 @@ nfp_bpf_check_exit(struct nfp_prog *nfp_prog,
 }
 
 static int
-nfp_bpf_check_ctx_ptr(struct nfp_prog *nfp_prog,
-		      const struct bpf_verifier_env *env, u8 reg)
+nfp_bpf_check_ptr(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+		  const struct bpf_verifier_env *env, u8 reg)
 {
-	if (env->cur_state.regs[reg].type != PTR_TO_CTX)
+	if (env->cur_state.regs[reg].type != PTR_TO_CTX &&
+	    env->cur_state.regs[reg].type != PTR_TO_PACKET)
 		return -EINVAL;
 
+	if (meta->ptr.type != NOT_INIT &&
+	    meta->ptr.type != env->cur_state.regs[reg].type)
+		return -EINVAL;
+
+	meta->ptr = env->cur_state.regs[reg];
+
 	return 0;
 }
 
@@ -145,11 +152,11 @@ nfp_verify_insn(struct bpf_verifier_env *env, int insn_idx, int prev_insn_idx)
 		return nfp_bpf_check_exit(priv->prog, env);
 
 	if ((meta->insn.code & ~BPF_SIZE_MASK) == (BPF_LDX | BPF_MEM))
-		return nfp_bpf_check_ctx_ptr(priv->prog, env,
-					     meta->insn.src_reg);
+		return nfp_bpf_check_ptr(priv->prog, meta, env,
+					 meta->insn.src_reg);
 	if ((meta->insn.code & ~BPF_SIZE_MASK) == (BPF_STX | BPF_MEM))
-		return nfp_bpf_check_ctx_ptr(priv->prog, env,
-					     meta->insn.dst_reg);
+		return nfp_bpf_check_ptr(priv->prog, meta, env,
+					 meta->insn.dst_reg);
 
 	return 0;
 }
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_asm.c b/drivers/net/ethernet/netronome/nfp/nfp_asm.c
index de76e7444fc2..7cae99b3e00a 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_asm.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_asm.c
@@ -42,6 +42,9 @@
 const struct cmd_tgt_act cmd_tgt_act[__CMD_TGT_MAP_SIZE] = {
 	[CMD_TGT_WRITE8] =		{ 0x00, 0x42 },
 	[CMD_TGT_READ8] =		{ 0x01, 0x43 },
+	[CMD_TGT_READ32] =		{ 0x00, 0x5c },
+	[CMD_TGT_READ32_LE] =		{ 0x01, 0x5c },
+	[CMD_TGT_READ32_SWAP] =		{ 0x02, 0x5c },
 	[CMD_TGT_READ_LE] =		{ 0x01, 0x40 },
 	[CMD_TGT_READ_SWAP_LE] =	{ 0x03, 0x40 },
 };
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_asm.h b/drivers/net/ethernet/netronome/nfp/nfp_asm.h
index c4c18dd5630a..e3df7a26724f 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_asm.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_asm.h
@@ -153,6 +153,7 @@ enum shf_op {
 
 enum shf_sc {
 	SHF_SC_R_ROT = 0,
+	SHF_SC_NONE = SHF_SC_R_ROT,
 	SHF_SC_R_SHF = 1,
 	SHF_SC_L_SHF = 2,
 	SHF_SC_R_DSHF = 3,
@@ -217,6 +218,9 @@ struct cmd_tgt_act {
 enum cmd_tgt_map {
 	CMD_TGT_READ8,
 	CMD_TGT_WRITE8,
+	CMD_TGT_READ32,
+	CMD_TGT_READ32_LE,
+	CMD_TGT_READ32_SWAP,
 	CMD_TGT_READ_LE,
 	CMD_TGT_READ_SWAP_LE,
 	__CMD_TGT_MAP_SIZE,
-- 
2.14.1

[PATCH net-next 11/12] nfp: bpf: direct packet access - write

[Jakub Kicinski]

This patch adds ability to write packet contents using pre-validated
packet pointers (direct packet access).

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 114 +++++++++++++++++++++++++--
 drivers/net/ethernet/netronome/nfp/nfp_asm.c |   2 +-
 drivers/net/ethernet/netronome/nfp/nfp_asm.h |   2 +-
 3 files changed, 109 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 975d63fbc1d5..139a4ebdc774 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -607,6 +607,35 @@ static int construct_data_ld(struct nfp_prog *nfp_prog, u16 offset, u8 size)
 	return data_ld(nfp_prog, tmp_reg, 0, size);
 }
 
+static int
+data_stx_host_order(struct nfp_prog *nfp_prog, u8 dst_gpr, swreg offset,
+		    u8 src_gpr, u8 size)
+{
+	unsigned int i;
+
+	for (i = 0; i * 4 < size; i++)
+		wrp_mov(nfp_prog, reg_xfer(i), reg_a(src_gpr + i));
+
+	emit_cmd(nfp_prog, CMD_TGT_WRITE8_SWAP, CMD_MODE_32b, 0,
+		 reg_a(dst_gpr), offset, size - 1, true);
+
+	return 0;
+}
+
+static int
+data_st_host_order(struct nfp_prog *nfp_prog, u8 dst_gpr, swreg offset,
+		   u64 imm, u8 size)
+{
+	wrp_immed(nfp_prog, reg_xfer(0), imm);
+	if (size == 8)
+		wrp_immed(nfp_prog, reg_xfer(1), imm >> 32);
+
+	emit_cmd(nfp_prog, CMD_TGT_WRITE8_SWAP, CMD_MODE_32b, 0,
+		 reg_a(dst_gpr), offset, size - 1, true);
+
+	return 0;
+}
+
 static void
 wrp_alu_imm(struct nfp_prog *nfp_prog, u8 dst, enum alu_op alu_op, u32 imm)
 {
@@ -1196,24 +1225,88 @@ static int mem_ldx8(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 	return mem_ldx(nfp_prog, meta, 8);
 }
 
-static int mem_stx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+static int
+mem_st_data(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	    unsigned int size)
+{
+	u64 imm = meta->insn.imm; /* sign extend */
+	swreg off_reg;
+
+	off_reg = re_load_imm_any(nfp_prog, meta->insn.off, imm_b(nfp_prog));
+
+	return data_st_host_order(nfp_prog, meta->insn.dst_reg * 2, off_reg,
+				  imm, size);
+}
+
+static int mem_st(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+		  unsigned int size)
 {
+	if (meta->ptr.type == PTR_TO_PACKET)
+		return mem_st_data(nfp_prog, meta, size);
+
 	return -EOPNOTSUPP;
 }
 
-static int mem_stx4_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+static int mem_st1(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_st(nfp_prog, meta, 1);
+}
+
+static int mem_st2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_st(nfp_prog, meta, 2);
+}
+
+static int mem_st4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_st(nfp_prog, meta, 4);
+}
+
+static int mem_st8(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
+	return mem_st(nfp_prog, meta, 8);
+}
+
+static int
+mem_stx_data(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	     unsigned int size)
+{
+	swreg off_reg;
+
+	off_reg = re_load_imm_any(nfp_prog, meta->insn.off, imm_b(nfp_prog));
+
+	return data_stx_host_order(nfp_prog, meta->insn.dst_reg * 2, off_reg,
+				   meta->insn.src_reg * 2, size);
+}
+
+static int
+mem_stx(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	unsigned int size)
+{
+	if (meta->ptr.type == PTR_TO_PACKET)
+		return mem_stx_data(nfp_prog, meta, size);
+
 	return -EOPNOTSUPP;
 }
 
+static int mem_stx1(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_stx(nfp_prog, meta, 1);
+}
+
+static int mem_stx2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_stx(nfp_prog, meta, 2);
+}
+
 static int mem_stx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
-	if (meta->ptr.type == PTR_TO_PACKET)
-		return -EOPNOTSUPP;
+	return mem_stx(nfp_prog, meta, 4);
+}
 
-	if (nfp_prog->act == NN_ACT_XDP)
-		return mem_stx4_xdp(nfp_prog, meta);
-	return mem_stx4_skb(nfp_prog, meta);
+static int mem_stx8(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return mem_stx(nfp_prog, meta, 8);
 }
 
 static int jump(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
@@ -1432,7 +1525,14 @@ static const instr_cb_t instr_cb[256] = {
 	[BPF_LDX | BPF_MEM | BPF_H] =	mem_ldx2,
 	[BPF_LDX | BPF_MEM | BPF_W] =	mem_ldx4,
 	[BPF_LDX | BPF_MEM | BPF_DW] =	mem_ldx8,
+	[BPF_STX | BPF_MEM | BPF_B] =	mem_stx1,
+	[BPF_STX | BPF_MEM | BPF_H] =	mem_stx2,
 	[BPF_STX | BPF_MEM | BPF_W] =	mem_stx4,
+	[BPF_STX | BPF_MEM | BPF_DW] =	mem_stx8,
+	[BPF_ST | BPF_MEM | BPF_B] =	mem_st1,
+	[BPF_ST | BPF_MEM | BPF_H] =	mem_st2,
+	[BPF_ST | BPF_MEM | BPF_W] =	mem_st4,
+	[BPF_ST | BPF_MEM | BPF_DW] =	mem_st8,
 	[BPF_JMP | BPF_JA | BPF_K] =	jump,
 	[BPF_JMP | BPF_JEQ | BPF_K] =	jeq_imm,
 	[BPF_JMP | BPF_JGT | BPF_K] =	jgt_imm,
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_asm.c b/drivers/net/ethernet/netronome/nfp/nfp_asm.c
index 7cae99b3e00a..830f6de25f47 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_asm.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_asm.c
@@ -40,7 +40,7 @@
 #include "nfp_asm.h"
 
 const struct cmd_tgt_act cmd_tgt_act[__CMD_TGT_MAP_SIZE] = {
-	[CMD_TGT_WRITE8] =		{ 0x00, 0x42 },
+	[CMD_TGT_WRITE8_SWAP] =		{ 0x02, 0x42 },
 	[CMD_TGT_READ8] =		{ 0x01, 0x43 },
 	[CMD_TGT_READ32] =		{ 0x00, 0x5c },
 	[CMD_TGT_READ32_LE] =		{ 0x01, 0x5c },
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_asm.h b/drivers/net/ethernet/netronome/nfp/nfp_asm.h
index e3df7a26724f..c26aa7e4a839 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_asm.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_asm.h
@@ -217,7 +217,7 @@ struct cmd_tgt_act {
 
 enum cmd_tgt_map {
 	CMD_TGT_READ8,
-	CMD_TGT_WRITE8,
+	CMD_TGT_WRITE8_SWAP,
 	CMD_TGT_READ32,
 	CMD_TGT_READ32_LE,
 	CMD_TGT_READ32_SWAP,
-- 
2.14.1

[PATCH net-next 12/12] nfp: bpf: support direct packet access in TC

[Jakub Kicinski]

Add support for direct packet access in TC, note that because
writing the packet will cause the verifier to generate a csum
fixup prologue we won't be able to offload packet writes from
TC, just yet, only the reads will work.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
 drivers/net/ethernet/netronome/nfp/bpf/jit.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/bpf/jit.c b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
index 139a4ebdc774..23fb11a41cc4 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/jit.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/jit.c
@@ -1135,12 +1135,25 @@ static int data_ind_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 static int mem_ldx_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
 		       u8 size)
 {
+	swreg dst = reg_both(meta->insn.dst_reg * 2);
+
 	switch (meta->insn.off) {
 	case offsetof(struct sk_buff, len):
 		if (size != FIELD_SIZEOF(struct sk_buff, len))
 			return -EOPNOTSUPP;
-		wrp_mov(nfp_prog,
-			reg_both(meta->insn.dst_reg * 2), plen_reg(nfp_prog));
+		wrp_mov(nfp_prog, dst, plen_reg(nfp_prog));
+		break;
+	case offsetof(struct sk_buff, data):
+		if (size != sizeof(void *))
+			return -EOPNOTSUPP;
+		wrp_mov(nfp_prog, dst, pptr_reg(nfp_prog));
+		break;
+	case offsetof(struct sk_buff, cb) +
+	     offsetof(struct bpf_skb_data_end, data_end):
+		if (size != sizeof(void *))
+			return -EOPNOTSUPP;
+		emit_alu(nfp_prog, dst,
+			 plen_reg(nfp_prog), ALU_OP_ADD, pptr_reg(nfp_prog));
 		break;
 	default:
 		return -EOPNOTSUPP;
-- 
2.14.1

Re: [PATCH net-next 00/12] nfp: bpf: support direct packet access

[David Miller]

From: Jakub Kicinski <jakub.kicinski@netronome.com>
Date: Thu, 12 Oct 2017 10:34:06 -0700

> The core of this series is direct packet access support.  With a
> small change to the verifier, the offloaded code can now make
> use of DPA.  We need to be careful to use kernel (after initial
> translation) offsets in our JIT.  Direct packet access also brings
> us to the problem of eBPF endianness.  After considering the 
> changes necessary we decided to not support translation on both
> BE and LE hosts, for now.
> 
> This series contains two fixes - one for compare instructions and
> one for ineffective jne optimization.  I chose to include fixes
> in this set because the code in -net works only with unreleased
> PoC FW (ABI version 1) and therefore nobody outside of Netronome
> can exercise it anyway.

Series applied, thank you.