Message-ID: <f7c6b5ed-a812-4f46-b40c-74d55e47f24c@amd.com>
Date: Tue, 26 Aug 2025 13:53:09 -0500
From: "Moger, Babu" <babu.moger@....com>
To: Reinette Chatre <reinette.chatre@...el.com>, corbet@....net,
 tony.luck@...el.com, Dave.Martin@....com, james.morse@....com,
 tglx@...utronix.de, mingo@...hat.com, bp@...en8.de,
 dave.hansen@...ux.intel.com
Cc: x86@...nel.org, hpa@...or.com, akpm@...ux-foundation.org,
 paulmck@...nel.org, rostedt@...dmis.org, Neeraj.Upadhyay@....com,
 david@...hat.com, arnd@...db.de, fvdl@...gle.com, seanjc@...gle.com,
 thomas.lendacky@....com, pawan.kumar.gupta@...ux.intel.com,
 yosry.ahmed@...ux.dev, sohil.mehta@...el.com, xin@...or.com,
 kai.huang@...el.com, xiaoyao.li@...el.com, peterz@...radead.org,
 me@...aill.net, mario.limonciello@....com, xin3.li@...el.com,
 ebiggers@...gle.com, ak@...ux.intel.com, chang.seok.bae@...el.com,
 andrew.cooper3@...rix.com, perry.yuan@....com, linux-doc@...r.kernel.org,
 linux-kernel@...r.kernel.org, manali.shukla@....com, gautham.shenoy@....com
Subject: Re: [PATCH v8 09/10] fs/resctrl: Introduce interface to modify
 io_alloc Capacity Bit Masks

Hi Reinette,

On 8/7/25 20:53, Reinette Chatre wrote:
> Hi Babu,
> 
> On 8/5/25 4:30 PM, Babu Moger wrote:
>> "io_alloc" feature enables direct insertion of data from I/O devices into
>> the cache. By directly caching data from I/O devices rather than first
>> storing the I/O data in DRAM, it reduces the demands on DRAM bandwidth and
>> reduces latency to the processor consuming the I/O data.
>>
>> When CDP is enabled, io_alloc routes traffic using the highest CLOSID
>> associated with the L3CODE resource. To ensure consistent cache allocation
>> behavior, the L3CODE and L3DATA resources must remain synchronized.
>> rdtgroup_init_cat() function takes both L3CODE and L3DATA into account when
>> initializing CBMs for new groups.  The io_alloc feature must maintain the
>> same behavior, ensuring that the Cache Bit Masks (CBMs) for both L3CODE and
>> L3DATA are updated together.
> 
> Please rework this copy&pasted text also and make specific to what this
> patch actually does.

Sure.

> 
>> Enable users to modify io_alloc CBMs (Capacity Bit Masks) via the
>> io_alloc_cbm resctrl file when io_alloc is enabled.
> 
> Here the changelog can provide overview of what is done by this patch
> when a user provides a new CBM. This can include that a CBM provided
> to CDP enabled resource will copy the CBM to the CDP peer.

Sure.

> 
>>
>> Signed-off-by: Babu Moger <babu.moger@....com>
>> ---
> 
> 
> 
>> ---
>>  Documentation/filesystems/resctrl.rst |  8 +++
>>  fs/resctrl/ctrlmondata.c              | 97 +++++++++++++++++++++++++++
>>  fs/resctrl/internal.h                 |  3 +
>>  fs/resctrl/rdtgroup.c                 |  3 +-
>>  4 files changed, 110 insertions(+), 1 deletion(-)
>>
>> diff --git a/Documentation/filesystems/resctrl.rst b/Documentation/filesystems/resctrl.rst
>> index 3002f7fdb2fe..d955e8525af0 100644
>> --- a/Documentation/filesystems/resctrl.rst
>> +++ b/Documentation/filesystems/resctrl.rst
>> @@ -187,6 +187,14 @@ related to allocation:
>>  			# cat /sys/fs/resctrl/info/L3/io_alloc_cbm
>>  			0=ffff;1=ffff
>>  
>> +		CBMs can be configured by writing to the interface.
>> +
>> +		Example::
>> +
>> +			# echo 1=FF > /sys/fs/resctrl/info/L3/io_alloc_cbm
> 
> It may be useful to demonstrate syntax when more than one CBM is modified.

Sure.

> 
>> +			# cat /sys/fs/resctrl/info/L3/io_alloc_cbm
>> +			0=ffff;1=00ff
>> +
>>  		When CDP is enabled "io_alloc_cbm" associated with the DATA and CODE
>>  		resources may reflect the same values. For example, values read from and
>>  		written to /sys/fs/resctrl/info/L3DATA/io_alloc_cbm may be reflected by
>> diff --git a/fs/resctrl/ctrlmondata.c b/fs/resctrl/ctrlmondata.c
>> index 641094aac322..1f69117f96f8 100644
>> --- a/fs/resctrl/ctrlmondata.c
>> +++ b/fs/resctrl/ctrlmondata.c
>> @@ -858,3 +858,100 @@ int resctrl_io_alloc_cbm_show(struct kernfs_open_file *of, struct seq_file *seq,
>>  	cpus_read_unlock();
>>  	return ret;
>>  }
>> +
>> +static int resctrl_io_alloc_parse_line(char *line,  struct rdt_resource *r,
>> +				       struct resctrl_schema *s, u32 closid)
>> +{
>> +	enum resctrl_conf_type peer_type;
>> +	struct resctrl_schema *peer_s;
>> +	struct rdt_parse_data data;
>> +	struct rdt_ctrl_domain *d;
>> +	char *dom = NULL, *id;
>> +	unsigned long dom_id;
>> +
>> +next:
>> +	if (!line || line[0] == '\0')
>> +		return 0;
>> +
>> +	dom = strsep(&line, ";");
>> +	id = strsep(&dom, "=");
>> +	if (!dom || kstrtoul(id, 10, &dom_id)) {
>> +		rdt_last_cmd_puts("Missing '=' or non-numeric domain\n");
>> +		return -EINVAL;
>> +	}
>> +
>> +	dom = strim(dom);
>> +	list_for_each_entry(d, &r->ctrl_domains, hdr.list) {
>> +		if (d->hdr.id == dom_id) {
>> +			data.buf = dom;
>> +			data.mode = RDT_MODE_SHAREABLE;
>> +			data.closid = closid;
>> +			if (parse_cbm(&data, s, d))
>> +				return -EINVAL;
>> +			/*
>> +			 * When CDP is enabled, update the schema for both CDP_DATA
>> +			 * and CDP_CODE.
>> +			 */
>> +			if (resctrl_arch_get_cdp_enabled(r->rid)) {
>> +				peer_type = resctrl_peer_type(s->conf_type);
>> +				peer_s = resctrl_get_schema(peer_type);
>> +				if (parse_cbm(&data, peer_s, d))
>> +					return -EINVAL;
> 
> CBM is still parsed twice. As I mentioned in v7 the parsing only need to be done
> once and then the resulting CBM can be copied to CDP peer.
> https://lore.kernel.org/lkml/82045638-2b26-4682-9374-1c3e400a580a@intel.com/

My bad.  Simplified now to copy the staged_config from source to its peer.

peer_type = resctrl_peer_type(s->conf_type);
memcpy(&d->staged_config[peer_type],
&d->staged_config[s->conf_type],
sizeof(*d->staged_config));



> 
>> +			}
>> +			goto next;
>> +		}
>> +	}
>> +
>> +	return -EINVAL;
>> +}
>> +
>> +ssize_t resctrl_io_alloc_cbm_write(struct kernfs_open_file *of, char *buf,
>> +				   size_t nbytes, loff_t off)
>> +{
>> +	struct resctrl_schema *s = rdt_kn_parent_priv(of->kn);
>> +	struct rdt_resource *r = s->res;
>> +	u32 io_alloc_closid;
>> +	int ret = 0;
>> +
>> +	/* Valid input requires a trailing newline */
>> +	if (nbytes == 0 || buf[nbytes - 1] != '\n')
>> +		return -EINVAL;
>> +
>> +	buf[nbytes - 1] = '\0';
>> +
>> +	cpus_read_lock();
>> +	mutex_lock(&rdtgroup_mutex);
>> +
>> +	rdt_last_cmd_clear();
>> +
>> +	if (!r->cache.io_alloc_capable) {
>> +		rdt_last_cmd_printf("io_alloc is not supported on %s\n", s->name);
>> +		ret = -ENODEV;
>> +		goto out_unlock;
>> +	}
>> +
>> +	rdt_last_cmd_clear();
> 
> Unnecessary rdt_last_cmd_clear().

Removed it.

> 
>> +	rdt_staged_configs_clear();
> 
> Placement of this can be improved by putting it closer to the code that
> populates the staged configs. That is, just before resctrl_io_alloc_parse_line().
> The flow is not symmetrical in that the out_unlock exit code always
> clears the buffer whether it was used or not. I think it will be easier to
> understand if the out_unlock *only* unlocks the locks and there is a new
> goto label, for example, "out_clear_configs" that calls rdt_staged_configs_clear()
> and is used after resctrl_io_alloc_parse_line().

Yes. Sound good.

> 
>> +
>> +	if (!resctrl_arch_get_io_alloc_enabled(r)) {
>> +		rdt_last_cmd_printf("io_alloc is not enabled on %s\n", s->name);
>> +		ret = -ENODEV;
>> +		goto out_unlock;
>> +	}
>> +
>> +	io_alloc_closid = resctrl_io_alloc_closid(r);
>> +
>> +	ret = resctrl_io_alloc_parse_line(buf, r, s, io_alloc_closid);
>> +
>> +	if (ret)
>> +		goto out_unlock;
>> +
>> +	ret = resctrl_arch_update_domains(r, io_alloc_closid);
>> +
>> +out_unlock:
>> +	rdt_staged_configs_clear();
>> +	mutex_unlock(&rdtgroup_mutex);
>> +	cpus_read_unlock();
>> +
>> +	return ret ?: nbytes;
>> +}
> 
> Reinette
> 

-- 
Thanks
Babu Moger

Powered by blists - more mailing lists