linux-kernel - Re: [PATCH V0 1/1] bootconfig: Increase max size of bootconfig from 32 KB to 256 KB for DCC support

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]

Message-ID: <9545ca51-ccda-64f0-bdd4-3b53e06785ad@quicinc.com>
Date:   Tue, 10 Jan 2023 20:38:54 +0530
From:   Souradeep Chowdhury <quic_schowdhu@...cinc.com>
To:     "Masami Hiramatsu (Google)" <mhiramat@...nel.org>
CC:     <linux-kernel@...r.kernel.org>,
        Bjorn Andersson <andersson@...nel.org>,
        <linux-arm-kernel@...ts.infradead.org>,
        <linux-arm-msm@...r.kernel.org>,
        "Sai Prakash Ranjan" <quic_saipraka@...cinc.com>,
        Sibi Sankar <quic_sibis@...cinc.com>,
        Rajendra Nayak <quic_rjendra@...cinc.com>
Subject: Re: [PATCH V0 1/1] bootconfig: Increase max size of bootconfig from
 32 KB to 256 KB for DCC support



On 1/10/2023 8:16 PM, Masami Hiramatsu (Google) wrote:
> On Tue, 10 Jan 2023 17:26:07 +0530
> Souradeep Chowdhury <quic_schowdhu@...cinc.com> wrote:
> 
>>
>>
>> On 1/9/2023 8:48 PM, Masami Hiramatsu (Google) wrote:
>>> On Mon, 9 Jan 2023 20:01:05 +0530
>>> Souradeep Chowdhury <quic_schowdhu@...cinc.com> wrote:
>>>
>>>> Increasing the memory size of bootconfig to be able to handle a max number of
>>>> 8192 nodes to be fitted in memory size of 256KB.
>>>
>>> Sorry, but you missed the 'xbc_node::data' stores the index of the data and
>>> that is uint16_t. So the XBC_DATA_MAX is fixed limitation.
>>>
>>> The number of nodes (XBC_NODE_MAX) can be expanded because I just decided it
>>> to keep the pre-compiled array size ~8KB. Maybe expanding it to 64KB just
>>> increase the size of kernel on init memory (and freed after boot).
>>>
>>> Could you tell me why you need such a big data for your DCC?
>>>
>>> Thank you,
>>
>> DCC is a debugging tool used in qcom which is needed to debug crashes
>> that can happen at boot-time. For debugging purposes a large number of
>> registers need to be configured in DCC driver which is to be fed via the
>> bootconfig file. For that we need to expand the nodes as well as memory
>> for using bootconfig.
> 
> Hmm, how many registers does DCC usually use? And how big the bootconfig
> file is usually? I have no idea about that.

So a typical bootconfig file for consumption of DCC looks like as follows

dcc_config {
         link_list_0 {
                 qcom-curr-link-list = 6
                 qcom-link-list = R_0x1781005c_1_apb,
                                  R_0x1782005c_1_apb
         }
         link_list_1 {
                 qcom-curr-link-list = 5
                 qcom-link-list = R_0x1784005c_1_apb
         }
}

The "qcom-link-list" field can have 1000s of register , based on that 
max nodes is increased to 8192.

> 
>> Can you let us know the changes that you suggest for doing the same? Is
>> it fine to just increase the XBC_NODE_MAX, do we also need to
>> change the uint16_t to u32 for proper storing of index values?
> 
> Expanding the number of max nodes is easy, just increase the XBC_NODE_MAX
> (must be less than 64k). That will also increase the memory consumption
> during the boot time even if the bootconfig is small. Anyway, it will be
> freed after boot, so it maybe OK.

So since the limit is 64K, 8192 is a valid value for max nodes.

> 
> But expanding the size of max bootconfig needs to change the type of
> the 'data' field to uint32_t (since that will be used for building
> bootconfig tool) and you also must confirm that `tools/bootconfig/bootconfig`
> can be built and pass the test-bootconfig.sh.
> Hmm, comparing with expanding the max number of XBC node, changing the
> 'data' type to uint32_t may not have much impact on memory consumption point
> of view, because it may increase only 20% of memory, but expanding the
> MAX_XBC_NODE always increases more than double.
> 
> Thus, if we can accept increasing the number of node, it should be OK to
> change the 'data' type.

That means from DCC point of view only increasing the max nodes is 
enough as increasing the data size is unrelated to increasing the max nodes?

> 
> BTW, I think now we don't need the ' __attribute__ ((__packed__))' for
> struct xbc_node. It was packed for reducing the size of array and able to
> pass 'compiled' bootconfig, but now it is just passed as a text data for
> safety.

> 
> Thank you,
> 
>>
>>
>>>
>>>>
>>>> Signed-off-by: Souradeep Chowdhury <quic_schowdhu@...cinc.com>
>>>> ---
>>>>    include/linux/bootconfig.h | 6 +++---
>>>>    1 file changed, 3 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/include/linux/bootconfig.h b/include/linux/bootconfig.h
>>>> index 1611f9d..64d233b 100644
>>>> --- a/include/linux/bootconfig.h
>>>> +++ b/include/linux/bootconfig.h
>>>> @@ -55,11 +55,11 @@ struct xbc_node {
>>>>    } __attribute__ ((__packed__));
>>>>    
>>>>    #define XBC_KEY		0
>>>> -#define XBC_VALUE	(1 << 15)
>>>> -/* Maximum size of boot config is 32KB - 1 */
>>>> +#define XBC_VALUE	(1 << 18)
>>>> +/* Maximum size of boot config is 256KB - 1 */
>>>>    #define XBC_DATA_MAX	(XBC_VALUE - 1)
>>>>    
>>>> -#define XBC_NODE_MAX	1024
>>>> +#define XBC_NODE_MAX	8192
>>>>    #define XBC_KEYLEN_MAX	256
>>>>    #define XBC_DEPTH_MAX	16
>>>>    
>>>> -- 
>>>> 2.7.4
>>>>
>>>
>>>
> 
>