linux-kernel - Re: [RFC PATCH 02/14] coresight: Introduce device access abstraction

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 30 Jul 2020 15:58:48 +0100
From:   Suzuki K Poulose <suzuki.poulose@....com>
To:     mathieu.poirier@...aro.org
Cc:     linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        mike.leach@...aro.org, coresight@...ts.linaro.org
Subject: Re: [RFC PATCH 02/14] coresight: Introduce device access abstraction

On 07/29/2020 08:56 PM, Mathieu Poirier wrote:
> On Wed, Jul 22, 2020 at 06:20:28PM +0100, Suzuki K Poulose wrote:
>> We are about to introduce support for sysreg access to ETMv4.4+
>> component. Since there are generic routines that access the
>> registers (e.g, CS_LOCK/UNLOCK , claim/disclaim operations, timeout)
>> and in order to preserve the logic of these operations at a single place
>> we introduce an abstraction layer for the accesses to a given device.
>> This will also be helpful in consolidating the sysfs.attribute helpers,
>> that we define per driver.
> 
> Please drop the last sentence, it doesn't add to the current patch.
> 

Sure.

>> diff --git a/drivers/hwtracing/coresight/coresight.c b/drivers/hwtracing/coresight/coresight.c
>> index e9c90f2de34a..38e9c03ab754 100644
>> --- a/drivers/hwtracing/coresight/coresight.c
>> +++ b/drivers/hwtracing/coresight/coresight.c
>> @@ -1387,6 +1387,54 @@ static int __init coresight_init(void)
>>   }

...

>>    * coresight_release_platform_data: Release references to the devices connected
>>    * to the output port of this device.
>> @@ -1451,6 +1499,7 @@ struct coresight_device *coresight_register(struct coresight_desc *desc)
>>   	csdev->type = desc->type;
>>   	csdev->subtype = desc->subtype;
>>   	csdev->ops = desc->ops;
>> +	csdev->access = desc->access;
>>   	csdev->orphan = false;
>>   
>>   	csdev->dev.type = &coresight_dev_type[desc->type];
>> diff --git a/include/linux/coresight.h b/include/linux/coresight.h
>> index 58fffdecdbfd..81ac708689f8 100644
>> --- a/include/linux/coresight.h
>> +++ b/include/linux/coresight.h
>> @@ -7,6 +7,7 @@
>>   #define _LINUX_CORESIGHT_H
>>   
>>   #include <linux/device.h>
>> +#include <linux/io.h>
>>   #include <linux/perf_event.h>
>>   #include <linux/sched.h>
>>   
>> @@ -114,6 +115,32 @@ struct coresight_platform_data {
>>   	struct coresight_connection *conns;
>>   };
>>   
>> +/**
>> + * struct csdev_access - Abstraction of a CoreSight device access.
>> + *
>> + * @no_iomem	: True if the device doesn't have iomem access.
>> + * @base	: When no_iomem == false, base address of the component
>> + * @read	: Read from the given "offset" of the given instance.
>> + * @write	: Write "val" to the given "offset".
>> + */
>> +struct csdev_access {
>> +	bool no_iomem;
> 
> I find the no_iomen to be difficult to understand, especially when prefixed with
> '!'.  Using "has_iomem" would be a lot more intuitive and would avoid extra
> mental gymnastics.


I agree. That was a bit of laziness in part, to limit the changes to the
existing drivers, where almost everyone, except the ETM would need to
simply use the MMIO approach. So, in order to keep those changes to
minimum, i.e, simply initialize the base, I used the inverted logic.
I will fix it.



>> +u32 coresight_relaxed_read32(struct coresight_device *csdev, u32 offset);
>> +u32 coresight_read32(struct coresight_device *csdev, u32 offset);
>> +void coresight_write32(struct coresight_device *csdev, u32 val, u32 offset);
>> +void coresight_relaxed_write32(struct coresight_device *csdev,
>> +			       u32 val,
>> +			       u32 offset);
>> +
> 
> Why are the 64 bit version outside of the #ifdef and the 32 bit within?

Mistake ;-). I will address all of the comments above in my next version.

...


>> +static inline u64 coresight_read64(struct coresight_device *csdev, u32 offset)
>> +{
>> +	WARN_ON_ONCE(1);
> 
> Not sure about the motivation behind using WARN_ON_ONCE(), and only in the read
> functions.  I would simply return 0 here.  After all if CONFIG_CORESIGHT is not
> defined they won't make it very far.
> 

If someone is reading the values, they might do something with the
value, i.e, make a decision when they shouldn't. This is just to prevent
such cases.

>> +	return 0;
>> +}
>> +
>> +static inline void coresight_relaxed_write64(struct coresight_device *csdev,
>> +					     u64 val,
>> +					     u32 offset)
>> +{
>> +}
>> +
>> +static inline void coresight_write64(struct coresight_device *csdev, u64 val, u32 offset)
>> +{
>> +}
>> +
>>   #endif
> 
> I will likely come back to this patch once I have reviewed the rest of the set.
> 

Sure. I am looking for thoughts on the proposed API (not ABI) changes,
as they are quite significant and invasive changes in the code, without
much functionality changes.


Thank you for the review !

Suzuki