Re: [PATCH/RFC bpf-next 02/16] bpf: refactor propagate_live implementation

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> On 27 Mar 2019, at 16:35, Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote:
> 
> On Tue, Mar 26, 2019 at 06:05:25PM +0000, Jiong Wang wrote:
>> Some code inside current implementation of "propagate_liveness" is a little
>> bit verbose.
>> 
>> This patch refactor them so the code looks more simple and more clear.
>> 
>> The redundant usage of "vparent->frame[vstate->curframe]" is removed as we
>> are here. It is safe to do this because "state_equal" has guaranteed that
>> vstate->curframe must be equal with vparent->curframe.
>> 
>> Signed-off-by: Jiong Wang <jiong.wang@xxxxxxxxxxxxx>
>> ---
>> kernel/bpf/verifier.c | 44 ++++++++++++++++++++++++++++++--------------
>> 1 file changed, 30 insertions(+), 14 deletions(-)
>> 
>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index 6cc8c38..245bb3c 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
>> @@ -6050,6 +6050,22 @@ static bool states_equal(struct bpf_verifier_env *env,
>> 	return true;
>> }
>> 
>> +static int propagate_liveness_reg(struct bpf_verifier_env *env,
>> +				  struct bpf_reg_state *reg,
>> +				  struct bpf_reg_state *parent_reg, u8 flag)
>> +{
>> +	int err;
>> +
>> +	if (parent_reg->live & flag || !(reg->live & flag))
>> +		return 0;
>> +
>> +	err = mark_reg_read(env, reg, parent_reg);
>> +	if (err)
>> +		return err;
>> +
>> +	return 1;
>> +}
> 
> what is the difference between 1 and 0 ? it doesn't seem to be used.

0 means no propagation has been done. 1 means propagation has been done.

They are used later in patch 4. If there is propagation, then will trigger
insn marking.

Will add comment for this.


> 
>> +
>> /* A write screens off any subsequent reads; but write marks come from the
>>  * straight-line code between a state and its parent.  When we arrive at an
>>  * equivalent state (jump target or such) we didn't arrive by the straight-line
>> @@ -6061,8 +6077,9 @@ static int propagate_liveness(struct bpf_verifier_env *env,
>> 			      const struct bpf_verifier_state *vstate,
>> 			      struct bpf_verifier_state *vparent)
>> {
>> -	int i, frame, err = 0;
>> +	struct bpf_reg_state *regs, *parent_regs;
>> 	struct bpf_func_state *state, *parent;
>> +	int i, frame, err = 0;
>> 
>> 	if (vparent->curframe != vstate->curframe) {
>> 		WARN(1, "propagate_live: parent frame %d current frame %d\n",
>> @@ -6071,16 +6088,13 @@ static int propagate_liveness(struct bpf_verifier_env *env,
>> 	}
>> 	/* Propagate read liveness of registers... */
>> 	BUILD_BUG_ON(BPF_REG_FP + 1 != MAX_BPF_REG);
>> +	parent_regs = vparent->frame[vparent->curframe]->regs;
>> +	regs = vstate->frame[vstate->curframe]->regs;
> 
> 
> may be do:
> frame = vstate->curframe;
> if (vparent->curframe != frame) { WARN...
> parent_regs = vparent->frame[frame]->regs;
> regs = vstate->frame[frame]->regs;
> 
> ?

Ack, will factor out "vstate->curframe” into “frame”.
 
And there is a check and warning on the equality already, just several lines above:

   if (vparent->curframe != vstate->curframe) {                             
     WARN(1, "propagate_live: parent frame %d current frame %d\n",

Regards,
Jiong
 
> 
>> 	/* We don't need to worry about FP liveness because it's read-only */
>> 	for (i = 0; i < BPF_REG_FP; i++) {
>> -		if (vparent->frame[vparent->curframe]->regs[i].live & REG_LIVE_READ)
>> -			continue;
>> -		if (vstate->frame[vstate->curframe]->regs[i].live & REG_LIVE_READ) {
>> -			err = mark_reg_read(env, &vstate->frame[vstate->curframe]->regs[i],
>> -					    &vparent->frame[vstate->curframe]->regs[i]);
>> -			if (err)
>> -				return err;
>> -		}
>> +		err = propagate_liveness_reg(env, &regs[i], &parent_regs[i]);
>> +		if (err < 0)
>> +			return err;
>> 	}
>> 
>> 	/* ... and stack slots */
>> @@ -6089,11 +6103,13 @@ static int propagate_liveness(struct bpf_verifier_env *env,
>> 		parent = vparent->frame[frame];
>> 		for (i = 0; i < state->allocated_stack / BPF_REG_SIZE &&
>> 			    i < parent->allocated_stack / BPF_REG_SIZE; i++) {
>> -			if (parent->stack[i].spilled_ptr.live & REG_LIVE_READ)
>> -				continue;
>> -			if (state->stack[i].spilled_ptr.live & REG_LIVE_READ)
>> -				mark_reg_read(env, &state->stack[i].spilled_ptr,
>> -					      &parent->stack[i].spilled_ptr);
>> +			struct bpf_reg_state *parent_reg, *reg;
>> +
>> +			parent_reg = &parent->stack[i].spilled_ptr;
>> +			reg = &state->stack[i].spilled_ptr;
>> +			err = propagate_liveness_reg(env, reg, parent_reg);
>> +			if (err < 0)
>> +				return err;
>> 		}
>> 	}
>> 	return err;
>> -- 
>> 2.7.4
>> 





[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux