* Re: BUG:af_packet fails to TX TSO frames
From: Willem de Bruijn @ 2017-10-12 17:25 UTC (permalink / raw)
To: Anton Ivanov; +Cc: Network Development, David Miller
In-Reply-To: <CAF=yD-+W2tyXqCS_ueZgu11BB8ON5hdKGOZQOWtfr+HfvqW=Sg@mail.gmail.com>
On Thu, Oct 12, 2017 at 12:30 PM, Willem de Bruijn
<willemdebruijn.kernel@gmail.com> wrote:
> On Thu, Oct 12, 2017 at 11:44 AM, Anton Ivanov
> <anton.ivanov@cambridgegreys.com> wrote:
>> Found it.
>>
>> Two bugs canceling each other.
>> The bind sequence in: psock_txring_vnet.c is wrong.
>>
>> It does the following addr.sll_protocol = htons(ETH_P_IP);
>> before calling bind.
>>
>> If you set addr.sll_protocol to ETH_P_ALL where it should have been in the
>> first place the test program blows up with -ENOBUFS
>
> There is no such requirement that the socket should bind to ETH_P_ALL.
>
>> I think what is happening is that this value is taken into account when
>> looking at "what should I use to segment it with" in skb_mac_gso_segment
>> which is invoked at the end of the verification chain which starts in
>> packet_direct_xmit in af_packet.c
>
> packet_snd sets skb->protocol based on the protocol that the packet
> socket is bound to. Binding to ETH_P_IP is the right choice here.
To avoid having to open multiple sockets for different protocols,
sockaddr_ll can also be passed in the msg_name argument on
each call.
^ permalink raw reply
* [PATCH net-next 00/12] nfp: bpf: support direct packet access
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, 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
^ permalink raw reply
* [PATCH net-next 02/12] nfp: bpf: reorder arguments to emit_ld_field_any()
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 01/12] bpf: verifier: set reg_type on context accesses in second pass
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski, alexei.starovoitov, daniel
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 03/12] nfp: bpf: add missing return in jne_imm optimization
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 04/12] nfp: bpf: fix compare instructions
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 06/12] nfp: bpf: implement byte swap instruction
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 05/12] nfp: bpf: add mov helper
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 07/12] nfp: bpf: support BPF offload only on little endian
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 08/12] nfp: bpf: fix context accesses
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 09/12] nfp: bpf: separate I/O from checks for legacy data load
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 10/12] nfp: bpf: add support for direct packet access - read
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 11/12] nfp: bpf: direct packet access - write
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH net-next 12/12] nfp: bpf: support direct packet access in TC
From: Jakub Kicinski @ 2017-10-12 17:34 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, Jakub Kicinski
In-Reply-To: <20171012173418.4029-1-jakub.kicinski@netronome.com>
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
^ permalink raw reply related
* [PATCH] Revert "net: qcom/emac: enforce DMA address restrictions"
From: Timur Tabi @ 2017-10-12 17:42 UTC (permalink / raw)
To: David S. Miller, netdev, David.Laight; +Cc: timur
This reverts commit df1ec1b9d0df57e96011f175418dc95b1af46821.
It turns out that memory allocated via dma_alloc_coherent is always
aligned to the size of the buffer, so there's no way the RRD and RFD
can ever be in separate 32-bit regions.
Signed-off-by: Timur Tabi <timur@codeaurora.org>
---
drivers/net/ethernet/qualcomm/emac/emac-mac.c | 39 +++++++++++----------------
1 file changed, 15 insertions(+), 24 deletions(-)
diff --git a/drivers/net/ethernet/qualcomm/emac/emac-mac.c b/drivers/net/ethernet/qualcomm/emac/emac-mac.c
index 0f5ece5d9507..9cbb27263742 100644
--- a/drivers/net/ethernet/qualcomm/emac/emac-mac.c
+++ b/drivers/net/ethernet/qualcomm/emac/emac-mac.c
@@ -734,11 +734,6 @@ static int emac_rx_descs_alloc(struct emac_adapter *adpt)
rx_q->rrd.size = rx_q->rrd.count * (adpt->rrd_size * 4);
rx_q->rfd.size = rx_q->rfd.count * (adpt->rfd_size * 4);
- /* Check if the RRD and RFD are aligned properly, and if not, adjust. */
- if (upper_32_bits(ring_header->dma_addr) !=
- upper_32_bits(ring_header->dma_addr + ALIGN(rx_q->rrd.size, 8)))
- ring_header->used = ALIGN(rx_q->rrd.size, 8);
-
rx_q->rrd.dma_addr = ring_header->dma_addr + ring_header->used;
rx_q->rrd.v_addr = ring_header->v_addr + ring_header->used;
ring_header->used += ALIGN(rx_q->rrd.size, 8);
@@ -772,18 +767,11 @@ int emac_mac_rx_tx_rings_alloc_all(struct emac_adapter *adpt)
/* Ring DMA buffer. Each ring may need up to 8 bytes for alignment,
* hence the additional padding bytes are allocated.
- *
- * Also double the memory allocated for the RRD so that we can
- * re-align it if necessary. The EMAC has a restriction that the
- * upper 32 bits of the base addresses for the RFD and RRD rings
- * must be the same. It is extremely unlikely that this is not the
- * case, since the rings are only a few KB in size. However, we
- * need to check for this anyway, and if the two rings are not
- * compliant, then we re-align.
*/
- ring_header->size = ALIGN(num_tx_descs * (adpt->tpd_size * 4), 8) +
- ALIGN(num_rx_descs * (adpt->rfd_size * 4), 8) +
- ALIGN(num_rx_descs * (adpt->rrd_size * 4), 8) * 2;
+ ring_header->size = num_tx_descs * (adpt->tpd_size * 4) +
+ num_rx_descs * (adpt->rfd_size * 4) +
+ num_rx_descs * (adpt->rrd_size * 4) +
+ 8 + 2 * 8; /* 8 byte per one Tx and two Rx rings */
ring_header->used = 0;
ring_header->v_addr = dma_zalloc_coherent(dev, ring_header->size,
@@ -792,23 +780,26 @@ int emac_mac_rx_tx_rings_alloc_all(struct emac_adapter *adpt)
if (!ring_header->v_addr)
return -ENOMEM;
- ret = emac_rx_descs_alloc(adpt);
- if (ret) {
- netdev_err(adpt->netdev, "error: Rx Queue alloc failed\n");
- goto err_alloc_rx;
- }
+ ring_header->used = ALIGN(ring_header->dma_addr, 8) -
+ ring_header->dma_addr;
ret = emac_tx_q_desc_alloc(adpt, &adpt->tx_q);
if (ret) {
- netdev_err(adpt->netdev, "transmit queue allocation failed\n");
+ netdev_err(adpt->netdev, "error: Tx Queue alloc failed\n");
goto err_alloc_tx;
}
+ ret = emac_rx_descs_alloc(adpt);
+ if (ret) {
+ netdev_err(adpt->netdev, "error: Rx Queue alloc failed\n");
+ goto err_alloc_rx;
+ }
+
return 0;
-err_alloc_tx:
- emac_rx_q_bufs_free(adpt);
err_alloc_rx:
+ emac_tx_q_bufs_free(adpt);
+err_alloc_tx:
dma_free_coherent(dev, ring_header->size,
ring_header->v_addr, ring_header->dma_addr);
--
Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm
Technologies, Inc. Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
^ permalink raw reply related
* Re: [next-queue PATCH v6 3/5] net/sched: Introduce Credit Based Shaper (CBS) qdisc
From: Vinicius Costa Gomes @ 2017-10-12 17:50 UTC (permalink / raw)
To: Eric Dumazet
Cc: netdev, intel-wired-lan, jhs, xiyou.wangcong, jiri, andre.guedes,
ivan.briano, jesus.sanchez-palencia, boon.leong.ong,
richardcochran, henrik, levipearson, rodney.cummings
In-Reply-To: <1507771972.31614.41.camel@edumazet-glaptop3.roam.corp.google.com>
Hi,
Eric Dumazet <eric.dumazet@gmail.com> writes:
> On Wed, 2017-10-11 at 17:54 -0700, Vinicius Costa Gomes wrote:
>> This queueing discipline implements the shaper algorithm defined by
>> the 802.1Q-2014 Section 8.6.8.2 and detailed in Annex L.
>
> ...
>
>> +static s64 delay_from_credits(s64 credits, s32 slope)
>> +{
>> + s64 rate = slope * BYTES_PER_KBIT;
>> +
>> + if (unlikely(rate == 0))
>> + return S64_MAX;
>> +
>> + return ((-credits * NSEC_PER_SEC) / rate);
>> +}
>
> Have you tried to compile this on 32bit arch ?
>
> make ARCH=i386
Will be fixed on v7.
Cheers,
^ permalink raw reply
* Re: [PATCH] Revert "net: qcom/emac: enforce DMA address restrictions"
From: David Miller @ 2017-10-12 17:50 UTC (permalink / raw)
To: timur; +Cc: netdev, David.Laight
In-Reply-To: <1507830124-19294-1-git-send-email-timur@codeaurora.org>
From: Timur Tabi <timur@codeaurora.org>
Date: Thu, 12 Oct 2017 12:42:04 -0500
> This reverts commit df1ec1b9d0df57e96011f175418dc95b1af46821.
>
> It turns out that memory allocated via dma_alloc_coherent is always
> aligned to the size of the buffer, so there's no way the RRD and RFD
> can ever be in separate 32-bit regions.
>
> Signed-off-by: Timur Tabi <timur@codeaurora.org>
Applied, thanks.
^ permalink raw reply
* Re: [Nouveau] [PATCH 03/10] driver:gpu: return -ENOMEM on allocation failure.
From: Daniel Vetter @ 2017-10-12 17:55 UTC (permalink / raw)
To: Allen Pais
Cc: linux-fbdev, linux-scsi, nouveau, megaraidlinux.pdl, linux-kernel,
dri-devel, target-devel, linux-crypto, netdev,
MPT-FusionLinux.pdl, linux-btrfs
In-Reply-To: <1505287939-14106-3-git-send-email-allen.lkml@gmail.com>
On Wed, Sep 13, 2017 at 01:02:12PM +0530, Allen Pais wrote:
> Signed-off-by: Allen Pais <allen.lkml@gmail.com>
Applied to drm-misc-next, thanks.
-Daniel
> ---
> drivers/gpu/drm/gma500/mid_bios.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/gma500/mid_bios.c b/drivers/gpu/drm/gma500/mid_bios.c
> index d75ecb3..1fa1633 100644
> --- a/drivers/gpu/drm/gma500/mid_bios.c
> +++ b/drivers/gpu/drm/gma500/mid_bios.c
> @@ -237,7 +237,7 @@ static int mid_get_vbt_data_r10(struct drm_psb_private *dev_priv, u32 addr)
>
> gct = kmalloc(sizeof(*gct) * vbt.panel_count, GFP_KERNEL);
> if (!gct)
> - return -1;
> + return -ENOMEM;
>
> gct_virtual = ioremap(addr + sizeof(vbt),
> sizeof(*gct) * vbt.panel_count);
> --
> 2.7.4
>
> _______________________________________________
> Nouveau mailing list
> Nouveau@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
^ permalink raw reply
* Re: BUG:af_packet fails to TX TSO frames
From: Anton Ivanov @ 2017-10-12 17:58 UTC (permalink / raw)
To: Willem de Bruijn; +Cc: Network Development, David Miller
In-Reply-To: <CAF=yD-KiMWMwvD-s=z4brqUrNbRe4-fRZAKupEc2FvVn0MQxuQ@mail.gmail.com>
On 10/12/17 18:25, Willem de Bruijn wrote:
> On Thu, Oct 12, 2017 at 12:30 PM, Willem de Bruijn
> <willemdebruijn.kernel@gmail.com> wrote:
>> On Thu, Oct 12, 2017 at 11:44 AM, Anton Ivanov
>> <anton.ivanov@cambridgegreys.com> wrote:
>>> Found it.
>>>
>>> Two bugs canceling each other.
>>> The bind sequence in: psock_txring_vnet.c is wrong.
>>>
>>> It does the following addr.sll_protocol = htons(ETH_P_IP);
>>> before calling bind.
>>>
>>> If you set addr.sll_protocol to ETH_P_ALL where it should have been in the
>>> first place the test program blows up with -ENOBUFS
>> There is no such requirement that the socket should bind to ETH_P_ALL.
There is no requirement to bind to ETH_P_IP either and most code
examples going back more than 10 years to the days of TCP Illustrated
use ALL.
I just did the v6 test - if you bind with ETH_P_IP GSO on TCPv6 is
broken and returns NOBUF and vice versa.
>>
>>> I think what is happening is that this value is taken into account when
>>> looking at "what should I use to segment it with" in skb_mac_gso_segment
>>> which is invoked at the end of the verification chain which starts in
>>> packet_direct_xmit in af_packet.c
>> packet_snd sets skb->protocol based on the protocol that the packet
>> socket is bound to. Binding to ETH_P_IP is the right choice here.
> To avoid having to open multiple sockets for different protocols,
> sockaddr_ll can also be passed in the msg_name argument on
> each call.
Does not work for vnet headers - it honors what you bound with. I tried
to bind with ETH_ALL and pass ETH_P_IP as an arg and it ENOBUF-ed
>
--
Anton R. Ivanov
Cambridge Greys Limited, England and Wales company No 10273661
http://www.cambridgegreys.com/
^ permalink raw reply
* Re: RFC(v2): Audit Kernel Container IDs
From: Eric W. Biederman @ 2017-10-12 17:59 UTC (permalink / raw)
To: Richard Guy Briggs
Cc: mszeredi-H+wXaHxf7aLQT0dZR+AlfA, trondmy-7I+n7zu2hftEKMMhf/gKZA,
Andy Lutomirski, jlayton-H+wXaHxf7aLQT0dZR+AlfA,
Carlos O'Donell, Linux API, Linux Containers, Paul Moore,
Linux Kernel, Eric Paris, Al Viro, David Howells, Linux Audit,
Simo Sorce, Linux Network Development, Linux FS Devel,
cgroups-u79uwXL29TY76Z2rM5mHXA, Steve Grubb
In-Reply-To: <20171012141359.saqdtnodwmbz33b2-bcJWsdo4jJjeVoXN4CMphl7TgLCtbB0G@public.gmane.org>
Richard Guy Briggs <rgb-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> writes:
> A namespace cannot directly migrate from one container to another but
> could be assigned to a newly spawned container. A namespace can be
> moved from one container to another indirectly by having that namespace
> used in a second process in another container and then ending all the
> processes in the first container.
Ugh no. The semantics here are way too mushy. We need a clean crisp
unambiguous definition or it will be impossible to get this correct and
impossible to use for any security purpose.
I understand the challenge. Some of the container managers share
namespaces between containers. Leading to things that are not really
contained.
Please make this concept like an indellibale die. Once you are stained
with it you can not escape. If you don't meet all of the criteria you
aren't stained.
The justification that I heard, and that seems legitimate is that it is
not timely and it is hard to make the connection between the distinct
unshare, setns, and clone events and what is happening in the kernel.
With that justification definitely the network namespace needs to be
stained if it is appropriate.
I also don't see why this can't be a special dedicated audit message.
I just looked at the code in the kernel and nlmsg_type is a u16. There
are only a handful of audit message types defined. There is absolutely
no reason to bring proc into this.
I have the same reservation as the others about defining a new cap for
this. It should be enough to make setting the container id a one time
thing for a set of processes and namespaces.
If this is going to be security it needs to be very simple and very well defined.
Eric
^ permalink raw reply
* Re: [PATCH net-next v2 1/1] bridge: return error code when deleting Vlan
From: Roman Mashak @ 2017-10-12 18:07 UTC (permalink / raw)
To: David Ahern
Cc: David Miller, Stephen Hemminger, Linux Kernel Network Developers
In-Reply-To: <fef6f65d-473c-ee8e-51cc-1382cb6f84b9@gmail.com>
On Thu, Oct 12, 2017 at 10:19 AM, David Ahern <dsahern@gmail.com> wrote:
> On 10/12/17 7:51 AM, Roman Mashak wrote:
>> v2:
>> Return err immediately if nbp_vlan_delete() fails (pointed by David Ahern)
>>
>> Signed-off-by: Roman Mashak <mrv@mojatatu.com>
>> ---
>> net/bridge/br_netlink.c | 8 +++++---
>> 1 file changed, 5 insertions(+), 3 deletions(-)
>>
>> diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
>> index f0e8268..1efdd48 100644
>> --- a/net/bridge/br_netlink.c
>> +++ b/net/bridge/br_netlink.c
>> @@ -527,11 +527,13 @@ static int br_vlan_info(struct net_bridge *br, struct net_bridge_port *p,
>>
>> case RTM_DELLINK:
>> if (p) {
>> - nbp_vlan_delete(p, vinfo->vid);
>> + err = nbp_vlan_delete(p, vinfo->vid);
>> + if (err)
>> + break;
>
> I'm not sure a break is the right thing to do. Seems like you leave it
> in a half configured state.
>
>> if (vinfo->flags & BRIDGE_VLAN_INFO_MASTER)
>> - br_vlan_delete(p->br, vinfo->vid);
>> + err = br_vlan_delete(p->br, vinfo->vid);
>> } else {
>> - br_vlan_delete(br, vinfo->vid);
>> + err = br_vlan_delete(br, vinfo->vid);
>> }
>> break;
>> }
>>
>
> Why do you want to return the error code here? Walking the code paths
> seems like ENOENT or err from switchdev_port_obj_del are the 2 error
> possibilities.
For example, if you attempt to delete a non-existing vlan on a port,
the current code succeeds and also sends event :
rtnetlink_rcv_msg
rtnl_bridge_dellink
br_dellink
br_afspec
br_vlan_info
int br_dellink(..)
{
...
err = br_afspec()
if (err == 0)
br_ifinfo_notify(RTM_NEWLINK, p);
}
This is misleading, so a proper errcode has to be produced.
^ permalink raw reply
* Re: [PATCH] [net] liquidio: fix timespec64_to_ns typo
From: Felix Manlunas @ 2017-10-12 18:07 UTC (permalink / raw)
To: Arnd Bergmann
Cc: Derek Chickles, Satanand Burla, Felix Manlunas, Raghu Vatsavayi,
David S. Miller, Intiyaz Basha, Rick Farrington, VSR Burru,
netdev, linux-kernel
In-Reply-To: <20171012094841.3178599-1-arnd@arndb.de>
On Thu, Oct 12, 2017 at 11:48:31AM +0200, Arnd Bergmann wrote:
> While experimenting with changes to the timekeeping code, I
> ran into a build error in the liquidio driver:
>
> drivers/net/ethernet/cavium/liquidio/lio_main.c: In function 'liquidio_ptp_settime':
> drivers/net/ethernet/cavium/liquidio/lio_main.c:1850:22: error: passing argument 1 of 'timespec_to_ns' from incompatible pointer type [-Werror=incompatible-pointer-types]
>
> The driver had a type mismatch since it was first merged, but
> this never caused problems because it is only built on 64-bit
> architectures that define timespec and timespec64 to the same
> type.
>
> If we ever want to compile-test the driver on 32-bit or change
> the way that 64-bit timespec64 is defined, we need to fix it,
> so let's just do it now.
>
> Fixes: f21fb3ed364b ("Add support of Cavium Liquidio ethernet adapters")
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
> drivers/net/ethernet/cavium/liquidio/lio_main.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/cavium/liquidio/lio_main.c b/drivers/net/ethernet/cavium/liquidio/lio_main.c
> index 963803bc6633..eafae3eb4fed 100644
> --- a/drivers/net/ethernet/cavium/liquidio/lio_main.c
> +++ b/drivers/net/ethernet/cavium/liquidio/lio_main.c
> @@ -1847,7 +1847,7 @@ static int liquidio_ptp_settime(struct ptp_clock_info *ptp,
> struct lio *lio = container_of(ptp, struct lio, ptp_info);
> struct octeon_device *oct = (struct octeon_device *)lio->oct_dev;
>
> - ns = timespec_to_ns(ts);
> + ns = timespec64_to_ns(ts);
>
> spin_lock_irqsave(&lio->ptp_lock, flags);
> lio_pci_writeq(oct, ns, CN6XXX_MIO_PTP_CLOCK_HI);
> --
> 2.9.0
>
Thanks.
Acked-by: Felix Manlunas <felix.manlunas@cavium.com>
^ permalink raw reply
* Re: Regression in throughput between kvm guests over virtual bridge
From: Wei Xu @ 2017-10-12 18:31 UTC (permalink / raw)
To: Matthew Rosato; +Cc: Jason Wang, netdev, davem, mst
In-Reply-To: <038445a6-9dd5-30c2-aac0-ab5efbfa7024@linux.vnet.ibm.com>
On Thu, Oct 05, 2017 at 04:07:45PM -0400, Matthew Rosato wrote:
>
> Ping... Jason, any other ideas or suggestions?
Hi Matthew,
Recently I am doing similar test on x86 for this patch, here are some,
differences between our testbeds.
1. It is nice you have got improvement with 50+ instances(or connections here?)
which would be quite helpful to address the issue, also you've figured out the
cost(wait/wakeup), kindly reminder did you pin uperf client/server along the whole
path besides vhost and vcpu threads?
2. It might be useful to short the traffic path as a reference, What I am running
is briefly like:
pktgen(host kernel) -> tap(x) -> guest(DPDK testpmd)
The bridge driver(br_forward(), etc) might impact performance due to my personal
experience, so eventually I settled down with this simplified testbed which fully
isolates the traffic from both userspace and host kernel stack(1 and 50 instances,
bridge driver, etc), therefore reduces potential interferences.
The down side of this is that it needs DPDK support in guest, has this ever be
run on s390x guest? An alternative approach is to directly run XDP drop on
virtio-net nic in guest, while this requires compiling XDP inside guest which needs
a newer distro(Fedora 25+ in my case or Ubuntu 16.10, not sure).
3. BTW, did you enable hugepage for your guest? It would performance more
or less depends on the memory demand when generating traffic, I didn't see
similar command lines in yours.
Hope this doesn't make it more complicated for you.:) We will keep working on this
and update you.
Thanks,
Wei
>
^ permalink raw reply
* Re: [PATCH net-next v2 1/1] bridge: return error code when deleting Vlan
From: Nikolay Aleksandrov @ 2017-10-12 18:12 UTC (permalink / raw)
To: Roman Mashak, David Ahern
Cc: David Miller, Stephen Hemminger, Linux Kernel Network Developers
In-Reply-To: <CAHvchGnVZaSq+s9953jdbf6-UzDFb36457U6Pk0oQHgVOKwzQg@mail.gmail.com>
On 12/10/17 21:07, Roman Mashak wrote:
> On Thu, Oct 12, 2017 at 10:19 AM, David Ahern <dsahern@gmail.com> wrote:
>> On 10/12/17 7:51 AM, Roman Mashak wrote:
>>> v2:
>>> Return err immediately if nbp_vlan_delete() fails (pointed by David Ahern)
>>>
>>> Signed-off-by: Roman Mashak <mrv@mojatatu.com>
>>> ---
>>> net/bridge/br_netlink.c | 8 +++++---
>>> 1 file changed, 5 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
>>> index f0e8268..1efdd48 100644
>>> --- a/net/bridge/br_netlink.c
>>> +++ b/net/bridge/br_netlink.c
>>> @@ -527,11 +527,13 @@ static int br_vlan_info(struct net_bridge *br, struct net_bridge_port *p,
>>>
>>> case RTM_DELLINK:
>>> if (p) {
>>> - nbp_vlan_delete(p, vinfo->vid);
>>> + err = nbp_vlan_delete(p, vinfo->vid);
>>> + if (err)
>>> + break;
>>
>> I'm not sure a break is the right thing to do. Seems like you leave it
>> in a half configured state.
>>
>>> if (vinfo->flags & BRIDGE_VLAN_INFO_MASTER)
>>> - br_vlan_delete(p->br, vinfo->vid);
>>> + err = br_vlan_delete(p->br, vinfo->vid);
>>> } else {
>>> - br_vlan_delete(br, vinfo->vid);
>>> + err = br_vlan_delete(br, vinfo->vid);
>>> }
>>> break;
>>> }
>>>
>>
>> Why do you want to return the error code here? Walking the code paths
>> seems like ENOENT or err from switchdev_port_obj_del are the 2 error
>> possibilities.
>
> For example, if you attempt to delete a non-existing vlan on a port,
> the current code succeeds and also sends event :
>
> rtnetlink_rcv_msg
> rtnl_bridge_dellink
> br_dellink
> br_afspec
> br_vlan_info
>
> int br_dellink(..)
> {
> ...
> err = br_afspec()
> if (err == 0)
> br_ifinfo_notify(RTM_NEWLINK, p);
> }
>
> This is misleading, so a proper errcode has to be produced.
>
True, but you also change the expected behaviour because now a user can
clear all vlans with one request (1 - 4094), and after the change that
will fail with a partial delete if some vlan was missing.
This has been the behaviour forever and some script might depend on it.
Also IMO, and as David also mentioned, doing a partial delete is not good.
^ permalink raw reply
* Re: [RFC 1/3] devlink: Add config parameter get/set operations
From: Andy Gospodarek @ 2017-10-12 18:12 UTC (permalink / raw)
To: Jiri Pirko
Cc: Steve Lin, netdev, jiri, davem, michael.chan, linux-pci, linville
In-Reply-To: <20171012140317.GC14672@nanopsycho>
On Thu, Oct 12, 2017 at 04:03:17PM +0200, Jiri Pirko wrote:
> Thu, Oct 12, 2017 at 03:34:20PM CEST, steven.lin1@broadcom.com wrote:
> >Add support for config parameter get/set commands. Initially used by
> >bnxt driver, but other drivers can use the same, or new, attributes.
> >The config_get() and config_set() operations operate as expected, but
> >note that the driver implementation of the config_set() operation can
> >indicate whether a restart is necessary for the setting to take
> >effect.
> >
>
> First of all, I like this approach.
>
> I would like to see this patch split into:
> 1) config-options infrastructure introduction
> 2) specific config options introductions - would be best to have it
> per-option. We need to make sure every option is very well described
> and explained usecases. This is needed in order vendors to share
> attributes among drivers.
>
> More nits inlined.
>
>
> >Signed-off-by: Steve Lin <steven.lin1@broadcom.com>
> >Acked-by: Andy Gospodarek <gospo@broadcom.com>
> >---
> > include/net/devlink.h | 4 +
> > include/uapi/linux/devlink.h | 108 ++++++++++++++++++++++
> > net/core/devlink.c | 207 +++++++++++++++++++++++++++++++++++++++++++
> > 3 files changed, 319 insertions(+)
> >
> > static inline void *devlink_priv(struct devlink *devlink)
> >diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
> >index 0cbca96..e959716 100644
> >--- a/include/uapi/linux/devlink.h
> >+++ b/include/uapi/linux/devlink.h
[...]
> >@@ -202,6 +267,49 @@ enum devlink_attr {
> >
> > DEVLINK_ATTR_ESWITCH_ENCAP_MODE, /* u8 */
> >
> >+ /* Configuration Parameters */
> >+ DEVLINK_ATTR_SRIOV_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_NUM_VF_PER_PF, /* u32 */
> >+ DEVLINK_ATTR_MAX_NUM_PF_MSIX_VECT, /* u32 */
> >+ DEVLINK_ATTR_MSIX_VECTORS_PER_VF, /* u32 */
> >+ DEVLINK_ATTR_NPAR_NUM_PARTITIONS_PER_PORT, /* u32 */
> >+ DEVLINK_ATTR_NPAR_BW_IN_PERCENT, /* u8 */
> >+ DEVLINK_ATTR_NPAR_BW_RESERVATION, /* u8 */
> >+ DEVLINK_ATTR_NPAR_BW_RESERVATION_VALID, /* u8 */
> >+ DEVLINK_ATTR_NPAR_BW_LIMIT, /* u8 */
> >+ DEVLINK_ATTR_NPAR_BW_LIMIT_VALID, /* u8 */
> >+ DEVLINK_ATTR_DCBX_MODE, /* u8 */
> >+ DEVLINK_ATTR_RDMA_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_MULTIFUNC_MODE, /* u8 */
> >+ DEVLINK_ATTR_SECURE_NIC_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_IGNORE_ARI_CAPABILITY, /* u8 */
> >+ DEVLINK_ATTR_LLDP_NEAREST_BRIDGE_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_LLDP_NEAREST_NONTPMR_BRIDGE_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_PME_CAPABILITY_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_MAGIC_PACKET_WOL_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_EEE_PWR_SAVE_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_AUTONEG_PROTOCOL, /* u8 */
> >+ DEVLINK_ATTR_MEDIA_AUTO_DETECT, /* u8 */
> >+ DEVLINK_ATTR_PHY_SELECT, /* u8 */
> >+ DEVLINK_ATTR_PRE_OS_LINK_SPEED_D0, /* u8 */
> >+ DEVLINK_ATTR_PRE_OS_LINK_SPEED_D3, /* u8 */
> >+ DEVLINK_ATTR_MBA_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_MBA_BOOT_TYPE, /* u8 */
> >+ DEVLINK_ATTR_MBA_DELAY_TIME, /* u32 */
> >+ DEVLINK_ATTR_MBA_SETUP_HOT_KEY, /* u8 */
> >+ DEVLINK_ATTR_MBA_HIDE_SETUP_PROMPT, /* u8 */
> >+ DEVLINK_ATTR_MBA_BOOT_RETRY_COUNT, /* u32 */
> >+ DEVLINK_ATTR_MBA_VLAN_ENABLED, /* u8 */
> >+ DEVLINK_ATTR_MBA_VLAN_TAG, /* u16 */
> >+ DEVLINK_ATTR_MBA_BOOT_PROTOCOL, /* u8 */
> >+ DEVLINK_ATTR_MBA_LINK_SPEED, /* u8 */
>
> Okay, I think it is about the time we should start thinking about
> putting this new config attributes under nester attribute. What do you
> think?
>
Steve and I actually had a similar discussion yesterday when I was doing
a final review of the patches.
My only objection to nesting was coming up with a way to describe these
functions that made them seem different than existing configuration
options. In this case of the hardware we are trying to support these
are all permanent config options, so we would call them
DEVLINK_ATTR_NVRAM or DEVLINK_ATTR_PERM. Does that seem reasonable to
others?
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox