Message-ID: <143e209f-c432-a587-0298-227c28875545@quicinc.com>
Date:   Tue, 25 Jul 2023 08:59:26 -0700
From:   Mike Tipton <quic_mdtipton@...cinc.com>
To:     Pavan Kondeti <quic_pkondeti@...cinc.com>
CC:     <quic_mdtipton@...cinc.com>, <djakov@...nel.org>,
        <gregkh@...uxfoundation.org>, <rafael@...nel.org>,
        <linux-kernel@...r.kernel.org>, <linux-pm@...r.kernel.org>,
        <quic_okukatla@...cinc.com>, <quic_viveka@...cinc.com>
Subject: Re: [PATCH 3/3] interconnect: Add debugfs test client

On 7/24/2023 10:46 PM, Pavan Kondeti wrote:
> On Mon, Jul 24, 2023 at 06:28:59PM -0700, Mike Tipton wrote:
>> It's often useful during test, debug, and development to issue path
>> votes from shell. Add a debugfs client for this purpose.
>>
>> Example usage:
>> 	cd /sys/kernel/debug/interconnect/test-client/
>>
>> 	# Configure node endpoints for the path from CPU to DDR on
>> 	# qcom/sm8550.
>> 	echo chm_apps > src_node
>> 	echo ebi > dst_node
>>
>> 	# Get path between src_node and dst_node. This is only
>> 	# necessary after updating the node endpoints.
>> 	echo 1 > get
>>
>> 	# Set desired BW to 1GBps avg and 2GBps peak.
>> 	echo 1000000 > avg_bw
>> 	echo 2000000 > peak_bw
>>
>> 	# Vote for avg_bw and peak_bw on the latest path from "get".
>> 	# Voting for multiple paths is possible by repeating this
>> 	# process for different nodes endpoints.
>> 	echo 1 > commit
>>
> 
> Can Documentation/driver-api/interconnect.rst be updated with this
> information?

Sure, will add some documentation there.

> 
>> Allowing userspace to directly enable and set bus rates can be dangerous
>> So, following in the footsteps of the regmap [0] and clk [1] frameworks,
>> keep these userspace controls compile-time disabled without Kconfig
>> options to enable them. Enabling this will require code changes to
>> define INTERCONNECT_ALLOW_WRITE_DEBUGFS.
>>   
> 
> Completely agree with your argument here.
> 
>> [0] commit 09c6ecd39410 ("regmap: Add support for writing to regmap registers via debugfs")
>> [1] commit 37215da5553e ("clk: Add support for setting clk_rate via debugfs")
>>
>> Signed-off-by: Mike Tipton <quic_mdtipton@...cinc.com>
>> ---
>>   drivers/interconnect/Makefile         |   2 +-
>>   drivers/interconnect/core.c           |   3 +
>>   drivers/interconnect/debugfs-client.c | 156 ++++++++++++++++++++++++++
>>   drivers/interconnect/internal.h       |   2 +
>>   4 files changed, 162 insertions(+), 1 deletion(-)
>>   create mode 100644 drivers/interconnect/debugfs-client.c
>>
>> diff --git a/drivers/interconnect/Makefile b/drivers/interconnect/Makefile
>> index 5604ce351a9f..d0888babb9a1 100644
>> --- a/drivers/interconnect/Makefile
>> +++ b/drivers/interconnect/Makefile
>> @@ -1,7 +1,7 @@
>>   # SPDX-License-Identifier: GPL-2.0
>>   
>>   CFLAGS_core.o				:= -I$(src)
>> -icc-core-objs				:= core.o bulk.o
>> +icc-core-objs				:= core.o bulk.o debugfs-client.o
>>   
>>   obj-$(CONFIG_INTERCONNECT)		+= icc-core.o
>>   obj-$(CONFIG_INTERCONNECT_IMX)		+= imx/
>> diff --git a/drivers/interconnect/core.c b/drivers/interconnect/core.c
>> index 511152398790..bc82571ff2a8 100644
>> --- a/drivers/interconnect/core.c
>> +++ b/drivers/interconnect/core.c
>> @@ -1117,6 +1117,9 @@ static int __init icc_init(void)
>>   			    icc_debugfs_dir, NULL, &icc_summary_fops);
>>   	debugfs_create_file("interconnect_graph", 0444,
>>   			    icc_debugfs_dir, NULL, &icc_graph_fops);
>> +
>> +	icc_debugfs_client_init(icc_debugfs_dir);
>> +
>>   	return 0;
>>   }
>>   
>> diff --git a/drivers/interconnect/debugfs-client.c b/drivers/interconnect/debugfs-client.c
>> new file mode 100644
>> index 000000000000..990dd2ff6df7
>> --- /dev/null
>> +++ b/drivers/interconnect/debugfs-client.c
>> @@ -0,0 +1,156 @@
>> +// SPDX-License-Identifier: GPL-2.0-only
>> +/*
>> + * Copyright (c) 2023, Qualcomm Innovation Center, Inc. All rights reserved.
>> + */
>> +#include <linux/debugfs.h>
>> +#include <linux/interconnect.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/uaccess.h>
>> +
> 
> Since you are not dealing with user copy directly, I wonder why this
> header inclusion is needed?

It's not. I'll remove it.

> 
>> +/*
>> + * This can be dangerous, therefore don't provide any real compile time
>> + * configuration option for this feature.
>> + * People who want to use this will need to modify the source code directly.
>> + */
>> +#undef INTERCONNECT_ALLOW_WRITE_DEBUGFS
>> +
> 
> Would it be nice to throw a compile error when DEBUG_FS is not enabled
> but INTERCONNECT_ALLOW_WRITE_DEBUGFS is enabled?

I don't think we want to consider that a compile error. Someone 
#defining this may still want the flexibility of disabling DEBUG_FS 
through Kconfig for certain configurations. Certain kernels could 
mainline their own changes to define INTERCONNECT_ALLOW_WRITE_DEBUGFS, 
and we don't want to preclude them from also disabling DEBUG_FS if desired.

> 
>> +#if defined(INTERCONNECT_ALLOW_WRITE_DEBUGFS) && defined(CONFIG_DEBUG_FS)
>> +
>> +static LIST_HEAD(debugfs_paths);
>> +static DEFINE_MUTEX(debugfs_lock);
>> +
>> +static struct platform_device *pdev;
>> +static struct icc_path *cur_path;
>> +
>> +static char *src_node;
>> +static char *dst_node;
>> +static u32 avg_bw;
>> +static u32 peak_bw;
>> +static u32 tag;
>> +
>> +struct debugfs_path {
>> +	const char *src;
>> +	const char *dst;
>> +	struct icc_path *path;
>> +	struct list_head list;
>> +};
>> +
>> +static struct icc_path *get_path(const char *src, const char *dst)
>> +{
>> +	struct debugfs_path *path;
>> +
>> +	list_for_each_entry(path, &debugfs_paths, list) {
>> +		if (!strcmp(path->src, src) && !strcmp(path->dst, dst))
>> +			return path->path;
>> +	}
>> +
>> +	return NULL;
>> +}
>> +
>> +static int icc_get_set(void *data, u64 val)
>> +{
>> +	struct debugfs_path *debugfs_path;
>> +	int ret = 0;
>> +
>> +	mutex_lock(&debugfs_lock);
>> +
>> +	/*
>> +	 * If we've already looked up a path, then use the existing one instead
>> +	 * of calling icc_get() again. This allows for updating previous BW
>> +	 * votes when "get" is written to multiple times for multiple paths.
>> +	 */
>> +	cur_path = get_path(src_node, dst_node);
>> +	if (cur_path)
>> +		goto out;
>> +
>> +	cur_path = icc_get(&pdev->dev, src_node, dst_node);
>> +	if (IS_ERR(cur_path)) {
>> +		ret = PTR_ERR(cur_path);
>> +		goto out;
>> +	}
> 
>  From the debugfs_write_file_str() implementation introduced in 1/3 of
> this series, these values (src_node/dst_node) needs RCU protection.
> Othewise, there is a potential UAF issue here, correct?

Yeah, good catch. This could happen if the "get" file is written to at 
the same time as the src_node or dst_node files. Will look into this.

> 
>> +
>> +	debugfs_path = kzalloc(sizeof(*debugfs_path), GFP_KERNEL);
>> +	if (!debugfs_path) {
>> +		ret = -ENOMEM;
>> +		goto err_put;
>> +	}
>> +
>> +	debugfs_path->path = cur_path;
>> +	debugfs_path->src = kstrdup(src_node, GFP_KERNEL);
>> +	debugfs_path->dst = kstrdup(dst_node, GFP_KERNEL);
>> +	if (!debugfs_path->src || !debugfs_path->dst) {
>> +		ret = -ENOMEM;
>> +		goto err_free;
>> +	}
>> +
>> +	list_add_tail(&debugfs_path->list, &debugfs_paths);
>> +	goto out;
>> +
>> +err_free:
>> +	kfree(debugfs_path->src);
>> +	kfree(debugfs_path->dst);
>> +	kfree(debugfs_path);
>> +err_put:
>> +	icc_put(cur_path);
>> +out:
>> +	mutex_unlock(&debugfs_lock);
>> +	return ret;
>> +}
>> +
>> +DEFINE_DEBUGFS_ATTRIBUTE(icc_get_fops, NULL, icc_get_set, "%llu\n");
>> +
>> +static int icc_commit_set(void *data, u64 val)
>> +{
>> +	int ret;
>> +
>> +	mutex_lock(&debugfs_lock);
>> +
>> +	if (IS_ERR_OR_NULL(cur_path)) {
>> +		ret = PTR_ERR(cur_path);
>> +		goto out;
>> +	}
>> +
>> +	icc_set_tag(cur_path, tag);
>> +	ret = icc_set_bw(cur_path, avg_bw, peak_bw);
>> +out:
>> +	mutex_unlock(&debugfs_lock);
>> +	return ret;
>> +}
>> +
>> +DEFINE_DEBUGFS_ATTRIBUTE(icc_commit_fops, NULL, icc_commit_set, "%llu\n");
>> +
>> +int icc_debugfs_client_init(struct dentry *icc_dir)
>> +{
>> +	struct dentry *client_dir;
>> +	int ret;
>> +
>> +	pdev = platform_device_alloc("icc-debugfs-client", PLATFORM_DEVID_AUTO);
>> +
> 
> In interconnect_summary debugfs file, it appears as
> "icc-debugfs-client.0.auto". Since we will only ever support one
> instance of this device, would PLATFORM_DEVID_NONE be more appropriate?

Agreed, no need for AUTO here. Will switch to NONE to make the device 
name prettier in the summary.

> 
>> +	ret = platform_device_add(pdev);
>> +	if (ret) {
>> +		pr_err("%s: failed to add platform device: %d\n", __func__, ret);
>> +		platform_device_put(pdev);
>> +		return ret;
>> +	}
>> +
>> +	client_dir = debugfs_create_dir("test_client", icc_dir);
>> +
>> +	debugfs_create_str("src_node", 0600, client_dir, &src_node);
>> +	debugfs_create_str("dst_node", 0600, client_dir, &dst_node);
>> +	debugfs_create_file("get", 0200, client_dir, NULL, &icc_get_fops);
>> +	debugfs_create_u32("avg_bw", 0600, client_dir, &avg_bw);
>> +	debugfs_create_u32("peak_bw", 0600, client_dir, &peak_bw);
>> +	debugfs_create_u32("tag", 0600, client_dir, &tag);
>> +	debugfs_create_file("commit", 0200, client_dir, NULL, &icc_commit_fops);
>> +
>> +	return 0;
>> +}
>> +
>> +#else
>> +
>> +int icc_debugfs_client_init(struct dentry *icc_dir)
>> +{
>> +	return 0;
>> +}
>> +
>> +#endif
>> diff --git a/drivers/interconnect/internal.h b/drivers/interconnect/internal.h
>> index f5f82a5c939e..87152e70a99b 100644
>> --- a/drivers/interconnect/internal.h
>> +++ b/drivers/interconnect/internal.h
>> @@ -41,4 +41,6 @@ struct icc_path {
>>   	struct icc_req reqs[];
>>   };
>>   
>> +int icc_debugfs_client_init(struct dentry *icc_dir);
>> +
>>   #endif
>> -- 
>> 2.17.1
>>
> 
> Thanks,
> Pavan

Powered by blists - more mailing lists