Re: [RESEND PATCH v7 4/5] ndctl/papr_scm,uapi: Add support for PAPR nvdimm specific methods

[Vaibhav Jain <vaibhav@linux.ibm.com>]

Thanks for reviewing this patch Aneesh.

"Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com> writes:

> Vaibhav Jain <vaibhav@linux.ibm.com> writes:
>
> ....
>
>  +
>> +/* Papr-scm-header + payload expected with ND_CMD_CALL ioctl from libnvdimm */
>> +struct nd_pdsm_cmd_pkg {
>> +	struct nd_cmd_pkg hdr;	/* Package header containing sub-cmd */
>> +	__s32 cmd_status;	/* Out: Sub-cmd status returned back */
>> +	__u16 payload_offset;	/* In: offset from start of struct */
>> +	__u16 payload_version;	/* In/Out: version of the payload */
>> +	__u8 payload[];		/* In/Out: Sub-cmd data buffer */
>> +} __packed;
>
> that payload_offset can be avoided if we prevent userspace to user a
> different variant of nd_pdsm_cmd_pkg which different header. We can keep
> things simpler if we can always find payload at
> nd_pdsm_cmd_pkg->payload.
Had introduced this member to handle case where new fields are added to
'struct nd_pdsm_cmd_pkg' without having to break the ABI. But agree with
the point that you made later that this can be simplified by replacing
'payload_offset' with a set of reserved variables. Will address this in
next iteration of this patchset.

>
>> +
>> +/*
>> + * Methods to be embedded in ND_CMD_CALL request. These are sent to the kernel
>> + * via 'nd_pdsm_cmd_pkg.hdr.nd_command' member of the ioctl struct
>> + */
>> +enum papr_scm_pdsm {
>> +	PAPR_SCM_PDSM_MIN = 0x0,
>> +	PAPR_SCM_PDSM_MAX,
>> +};
>> +
>> +/* Convert a libnvdimm nd_cmd_pkg to pdsm specific pkg */
>> +static inline struct nd_pdsm_cmd_pkg *nd_to_pdsm_cmd_pkg(struct nd_cmd_pkg *cmd)
>> +{
>> +	return (struct nd_pdsm_cmd_pkg *) cmd;
>> +}
>> +
>> +/* Return the payload pointer for a given pcmd */
>> +static inline void *pdsm_cmd_to_payload(struct nd_pdsm_cmd_pkg *pcmd)
>> +{
>> +	if (pcmd->hdr.nd_size_in == 0 && pcmd->hdr.nd_size_out == 0)
>> +		return NULL;
>> +	else
>> +		return (void *)((__u8 *) pcmd + pcmd->payload_offset);
>> +}
>> +
>
> we need to make sure userspace is not passing a wrong payload_offset.
Agree, that this function should have more strict checking for
payload_offset field. However will be getting rid of
'payload_offset' all together in the next iteration as you previously
suggested.

> and in the next patch you do
>
> +	/* Copy the health struct to the payload */
> +	memcpy(pdsm_cmd_to_payload(pkg), &p->health, copysize);
> +	pkg->hdr.nd_fw_size = copysize;
> +
Yes this is already being done in the patchset and changes proposed to
this pdsm_cmd_to_payload() should not impact other patches as
pdsm_cmd_to_payload() abstracts rest of the code from how to access the
payload.

> All this can be simplified if you can keep payload at
> nd_pdsm_cmd_pkg->payload.
>
> If you still want to have the ability to extend the header, then added a
> reserved field similar to nd_cmd_pkg.
>
Agree to this and will address this in V8.

>
> -aneesh

-- 
Cheers
~ Vaibhav