From: Hans de Goede [mailto:email@example.com]
Sent: Tuesday, November 27, 2018 11:21 AM
To: Rafael J. Wysocki <rafael(a)kernel.org>
Cc: Moore, Robert <robert.moore(a)intel.com>; Schmauss, Erik
<erik.schmauss(a)intel.com>; Rafael J. Wysocki <rjw(a)rjwysocki.net>; Len
Brown <lenb(a)kernel.org>; ACPI Devel Maling List <linux-
Subject: Re: [PATCH 4.20 regression fix] ACPICA: Fix handling of buffer-size in
On 27-11-18 17:40, Rafael J. Wysocki wrote:
> On Tue, Nov 27, 2018 at 9:39 AM Hans de Goede <hdegoede(a)redhat.com>
>> On 27-11-18 04:11, Moore, Robert wrote:
>>> Also, please send the acpidump for whatever machine is showing a
>>> We want to look closely at the surface 3 code vs. the code for this
>> Bob, I know you are unhappy with / frustrated about this part of the
>> ACPI specification (*), but please read the !@#$ patch description /
>> commit message.
>> The current problem has very little to do with the specification
>> being ambiguous, the code currently in 4.20 is using a struct field
>> which is
>> *clearly* marked in the specification as "reserved" as length
>> argument to a memcpy without checking that either the source or the
>> destination buffer is big enough.
>> Nor does it guarantee that we copy *enough* data when doing a write
>> to an i2c device through the GSB code, resulting in the OpRegion
>> handler writing whatever was in the tmp-buffer when we got it from
>> acpi_ut_create_buffer_object to the i2c-device, crashing the machine.
> The description of the problem is clear to me and and the patch does
> make it go away.
> Question is if the patch as is clashes with any future ACPICA work,
> because if it does, it may be better to do something else in
> principle. So far this is still unclear, but this regression needs to
> go away before final 4.20 and the patch still is there in my
> linux-next branch.
>> I can give you the acpidump of the non-booting device but it is not
>> doing anything special,
> Sending it over wouldn't hurt though IMO, would it?
True, here it is: https://fedorapeople.org/~jwrdegoede/acpidump.acer-
>> it is only using AttribByte accesses, the problem is not the DSDT,
>> the problem is the 4.20 code clearly, *obviously*, being buggy, so
>> please review the patch first.
>> *) IMHO this mostly is about the Surface line of products abusing the
>> specification, all other uses I've seen are very straight forward
> So until recently it has not been clear to Bob/Erik if there are any
> other production users of GenericSerialBus AFAICS and it would help to
> give them ASL with it from whatever machines you have access to.
Thanks for your patience
I do not think that is the case. A lot of laptops out there use the
GenericSerialBus interface in the form of an I2cSerialBus for Battery
monitoring, but all these laptops pretty much exclusively use
GenericSerialBus OpRegion accesses with AccessAs set to either AttribByte
or AttribBytes and that has been working fine for years and the spec is clear
enough on how to interpret these (esp. in combination with the DSTDs using
them which make the intention clear).
The problem with the Surface line of devices is that the use AccessAs
AttribRawProcessBytes on which the spec is a bit handwavy, the patch in 4.20
to fix AttribRawProcessBytes breaks AttribByte and AttribBytes accesses,
because for those the length field of the ACPI buffer passed when writing
the field is reserved, so we should not use it.
Thanks for this data point
Frankly given that using GenericSerialBus OpRegion accesses with AccessAs
set to either AttribByte or AttribBytes is very common, I'm amazed that there
have not been more regressions reported against 4.20. Note these do not
necessarily have to end up as system not booting, but could also e.g. be
battery monitoring behaving weirdly.
Note I was involved in the acpica discussion surrounding the patches to deal
with AttribRawProcessBytes, only I was so focused on
AttribRawProcessBytes that I failed to see they would break AttribByte and
AttribBytes accesses at that time, only to have it poetically bite myself in the
ass and at the end of a long bisect end up at the very patch I was sideways
We looked at the spec and your patch and we came to the following conclusions:
I will refer to the contents of the original patch using the # character to avoid
confusion with >. In the original patch you stated this:
#The ACPI 6.0 spec (ACPI_6.0.pdf) "126.96.36.199.5.2 Declaring and Using a
#GenericSerialBusData Buffer" (page 232) states that the GenericSerialBus Data
#Buffer Length field is only valid when doing a Read/Write Block (AttribBlock) transfer,
#but since the troublesome commit we unconditionally use the len field to determine
#how much data to copy from the source-buffer into the temp-buffer passed to the
The spec is wrong on this part. LEN is used in AttribBlock, AttribBlockProcessCall, and
AttribRawProcessBytes, we need to know how much data to copy. Len has to be
correct for the handler. Your solution is treating AttribBlock, AttribBlockProcessCall,
and AttribRawProcessBytes in the same way as everything else. I think we can try to
attempt to "honor" the spirit of these protocols by looking at LEN in certain
However, LEN is not always reliable.
#3) The temp-buffer passed to the OpRegion is allocated to the size returned by
#acpi_ex_get_serial_access_length(), which may be as little as 1, so potentially this
#may lead to a write overflow of the temp-buffer.
The temp buffer is at least 2. We missed adding 2 to the DataLengthField.
With the new approach, we let the temporary buffer be of the max size. Since LEN is
8 bits, we can only have a max data length of 256. Accounting for the STAT and LEN
fields, the Temp Buffer should be 258. In other words, we can't overflow the buffer
but we definitely need to choose wisely between the two.
Anyway aside from all of those details, the primary issue here is the DataLength
and also how to go about choosing the appropriate one. This will give the handler the
correct data to deal with.
There are hints in the spec that imply that LEN is important for some protocols but the
value of LEN cannot be trusted so we need to make sure to check to make sure that
this value is sane.
Could you modify your patch so that we take the DataLength as correct as possible?
Here's what I'm thinking:
AttribQuick - data length can be one or greater but with the structure of these buffers,
2 might be easier for the handler to use.
AttribSendRecieve - data length can be 3 (2 for STAT/ LEN header + 1 data)
AttribByte - datalength can be 3 (same as above)
AttribWord - datalength can be 4 (2 header + 2 data)
AttribBlock - datalength should be LEN unless LEN is greater than the source buffer.
In this case, DataLength should be the SourceBuffer's Length
or the max (whichever one is smaller)
AttribProcessCall - should be 4 (2 header + 2 data)
AttribBlockProcessCall - same as AttribBlock
AttribBytes - ignore AttribBytesFiedl and use LEN like AttribBlock
AttribRawBytes - same as AttribBytes (this one is confusing)
AtrribRawProcessBytes - same as AttribBytes
I think this would be a better solution. Let me know what you think and apologies for the
> I would suggest opening a BZ for that and attaching some relevant
> acpidump files in there, for future reference if nothing else.
> If you make a claim like the above, it's rather good to support it
> with some evidence. :-)
>>> -----Original Message-----
>>> From: Moore, Robert
>>> Sent: Monday, November 26, 2018 2:33 PM
>>> To: Schmauss, Erik <erik.schmauss(a)intel.com>; Rafael J. Wysocki
>>> <rjw(a)rjwysocki.net>; Hans de Goede <hdegoede(a)redhat.com>
>>> Cc: Len Brown <lenb(a)kernel.org>; linux-acpi(a)vger.kernel.org;
>>> Subject: RE: [PATCH 4.20 regression fix] ACPICA: Fix handling of
>>> buffer-size in acpi_ex_write_data_to_field()
>>> Looks like another round of issues with the completely ill-defined
generic serial bus. Below is a summary of the problems we have seen with
the Length field.
>>> I have not had time to completely read through this thread yet, so I'm
posting some text that I wrote a month or two ago.
>>> Let me say this, however:
>>> The biggest (and it is very big) issue is that the GenericSerialBus
interfaces are so poorly defined in the ACPI specification that any
implementation simply must guess at the proper behavior (Both the ACPI
implementation and any related drivers).
>>> Problems with these ACPI chapters:
>>> 188.8.131.52.6.2 Declaring and Using a GenericSerialBus Data Buffer
>>> 184.108.40.206.6.3 Using the GenericSerialBus Protocols
>>> 1) All of the ASL code examples do not compile due to syntax errors.
>>> There are
>>> 10 such examples, often with multiple syntax errors.
>>> 2) The basic/common data buffer format is not fully defined:
>>> typedef struct
>>> BYTE Status; // Byte 0 of the data buffer
>>> BYTE Length; // Byte 1 of the data buffer
>>> BYTE[x-1] Data; // Bytes 2-x of the arbitrary length data
>>> } // where x is the last index of the overall buffer
>>> The "Length" field is not defined. Is it the length of
"Data" or the Length
of the entire structure (Status + Length + Data)?
>>> The example code from the spec illustrates how the "Length" term
the structure above is undefined. Again, is it the "length of entire
or is it the "length of the Data element"?. The example code makes it appear
that it is the length of the entire structure, but it is not fully clear because it is
wrong in either case:
>>> Name(BUFF, Buffer(34)()) // Create GenericSerialBus data
>>> as BUFF
>>> CreateByteField(BUFF, 0x00, STAT) // STAT = Status (Byte)
>>> CreateByteField(BUFF, 0x01, LEN) // LEN = Length (Byte)
>>> CreateField(BUFF, 0x10, 256, DATA) // Data (Block)
>>> Store("ACPI", DATA) // Fill in outgoing data
>>> Store(8, LEN) // Length of the valid data
>>> (Actually should be 6? Or should it be 4?)