1083298 - Gfx driver block request: IntelGMAX4500HD

Status: RESOLVED INVALID

(Toolkit :: Blocklist Policy Requests, defect)

Description

[Nicolas Silva [:nical]]

Vendor: 0x8086
Devices: 
0x2a42, 0x2a43, 0x2e42, 0x2e43, 0x2e92, 0x2e93, 0x2e32, 0x2e33, 0x2e22, 0x2e23, 0x2e12, 0x2e13, 0x0042, 0x0046, 0x0102, 0x0106, 0x0112, 0x0116, 0x0122, 0x0126, 0x0166, 0x010a, 0x0080, 

Feature: ALL_FEATURES
Driver versions: anything < 8.15.10.2321

Reasons: Bug 1083071

Comment 1

[Milan Sreckovic [:milan] (needinfo for best results)]

Jorge, are you the right person to help with this?  See the e-mail thread for context.

Comment 2

[Jorge Villalobos [:jorgev] (he/him)]

Yes, I'll stage it shortly.

Comment 3

[Jorge Villalobos [:jorgev] (he/him)]

Do I need to set the Feature status to BLOCKED_DEVICE or anything else? How about the OS?

Comment 4

[Milan Sreckovic [:milan] (needinfo for best results)]

This is Windows only.  Don't know the answer to the Feature status question - Benoit?  Just accounting for the time difference, :nical probably can't answer before tomorrow.

Comment 5

[Benoit Jacob [:bjacob] (mostly away)]

The feature status doesn't matter very much. BLOCKED_DEVICE is fine here, since we're blocking specifically a particular device.

Comment 6

[Jorge Villalobos [:jorgev] (he/him)]

I created the block in staging. Please test: https://wiki.mozilla.org/Blocklisting/Testing

I set the OS to WINNT, though I noticed previous blocks targeted specific versions of WINNT, so I don't know if this will work.

Comment 7

[Milan Sreckovic [:milan] (needinfo for best results)]

Asked the submitter of bug 1083071 for testing if possible.

Comment 8

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

I tried to verify the blocklisting with the following scenario, on Win 7 x64, with Firefox 33:
1. Created "spoofed-firefox.bat" file with following content:
SET MOZ_GFX_SPOOF_WINDOWS_VERSION=60001
SET MOZ_GFX_SPOOF_VENDOR_ID=0x8086
SET MOZ_GFX_SPOOF_DEVICE_ID=0x0166
SET MOZ_GFX_SPOOF_DRIVER_VERSION=8.15.10.2320
"C:\Mozilla\Firefox\firefox.exe" -p -no-remote
2. Saved the file, opened it to launch Firefox, created a new profile, started Firefox, and verified "about:support" -> Graphics. ---> features were enabled as expected.
3. Went to about:config and set the following:
extensions.blocklist.url = "https://blocklist.addons-dev.allizom.org/...."
extensions.blocklist.itemURL = "https://blocklist.addons-dev.allizom.org/%LOCALE%/%APP%/blocked/%blockID%"
extensions.blocklist.detailsURL = "https://www.allizom.org/%LOCALE%/blocklist/"
extensions.blocklist.interval = 1
4. Closed Firefox and restarted it by running "spoofed-firefox.bat" and left it for a while.
5. Closed Firefox and restarted it by running "spoofed-firefox.bat".

Results:
- about:support still displayed features enabled (about_support_1.png). Features should have been blocked.
- about:config indicated that there were 2 pings and the last update was done at Thu, 16 Oct 2014 12:55:34 GMT (about_config_1.png).

As far as I can tell, the update of the blocklist was done, but versions smaller than 8.15.10.2321 were still getting blocked. I tried for various deviceIDs with the same result: blocklisting did not seem to have updated.

I'll continue investigating this, but initial results seem to show that this is not working.

Comment 9

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

Attachment: about_support_1.png

Comment 10

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

Attachment: about_config_1.png

Comment 11

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

Note that I've also tried forcing a ping as instructed in https://wiki.mozilla.org/Blocklisting/Testing, again got the data in about:config indicating that an update was made to the blocklist, but about:support still did not indicate any blocking for 8.15.10.2320. Trying values for DeviceID=0x2E23, shows that old values are still used: drivers up to 6.14.10.5218 have all features blocked, and versions up to 8.15.10.2202 have D2D blocked, and versions between 8.15.10.2202 and 8.15.10.2320 are still NOT blocked.

Note that we cannot test with real hardware as our graphic cards (HD Graphics 2500) do not support install of older drivers.

Comment 12

[Sylvestre Ledru [:Sylvestre]]

To make sure Jorge sees it

Comment 13

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

(In reply to Florin Mezei, QA (:FlorinMezei) from comment #8)
> As far as I can tell, the update of the blocklist was done, but versions
> smaller than 8.15.10.2321 were still getting blocked. I tried for various
> deviceIDs with the same result: blocklisting did not seem to have updated.

Small correction: "... versions smaller than 8.15.10.2321 were still NOT getting blocked."

Comment 14

[Jorge Villalobos [:jorgev] (he/him)]

Sorry, it looks like there were some changes in the URLs for the blocklist. I updated the docs:

> Go to about:config, find the extensions.blocklist.url pref and change the blocklist.addons.mozilla.org
> part of the value to blocklist-dev.allizom.org

Please test again making sure you use the right blolcklist-dev URL.

Comment 15

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

Tried with the new URL, following steps from comment 8, and separately exact steps from https://wiki.mozilla.org/Blocklisting/Testing, but still version 8.15.10.2320 is not blocked, and older values are still used for blocklisting.

Comment 16

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

Do you know what could be wrong Jorge?

Comment 17

[Jorge Villalobos [:jorgev] (he/him)]

Can you confirm that you have the following in your blocklist.xml?

    <gfxBlacklistEntry  blockID="g582">      <os>WINNT</os>      <vendor>0x8086</vendor>              <devices>
                      <device>0x2a42</device>
                      <device>0x2a43</device>
                      <device>0x2e42</device>
                      <device>0x2e43</device>
                      <device>0x2e92</device>
                      <device>0x2e93</device>
                      <device>0x2e32</device>
                      <device>0x2e33</device>
                      <device>0x2e22</device>
                      <device>0x2e23</device>
                      <device>0x2e12</device>
                      <device>0x2e13</device>
                      <device>0x0042</device>
                      <device>0x0046</device>
                      <device>0x0102</device>
                      <device>0x0106</device>
                      <device>0x0112</device>
                      <device>0x0116</device>
                      <device>0x0122</device>
                      <device>0x0126</device>
                      <device>0x0166</device>
                      <device>0x010a</device>
                      <device>0x0080</device>
                  </devices>
            <feature>ALL_FEATURES</feature>      <featureStatus>BLOCKED_DEVICE</featureStatus>      <driverVersion>8.15.10.2321</driverVersion>      <driverVersionComparator>LESS_THAN</driverVersionComparator>    </gfxBlacklistEntry>

Comment 18

[Tracy Walker [:tracy]]

(In reply to Jorge Villalobos [:jorgev] from comment #17)
> Can you confirm that you have the following in your blocklist.xml?
> 
>     <gfxBlacklistEntry  blockID="g582">      <os>WINNT</os>     
> <vendor>0x8086</vendor>              <devices>
>                       <device>0x2a42</device>
>                       <device>0x2a43</device>
>                       <device>0x2e42</device>
>                       <device>0x2e43</device>
>                       <device>0x2e92</device>
>                       <device>0x2e93</device>
>                       <device>0x2e32</device>
>                       <device>0x2e33</device>
>                       <device>0x2e22</device>
>                       <device>0x2e23</device>
>                       <device>0x2e12</device>
>                       <device>0x2e13</device>
>                       <device>0x0042</device>
>                       <device>0x0046</device>
>                       <device>0x0102</device>
>                       <device>0x0106</device>
>                       <device>0x0112</device>
>                       <device>0x0116</device>
>                       <device>0x0122</device>
>                       <device>0x0126</device>
>                       <device>0x0166</device>
>                       <device>0x010a</device>
>                       <device>0x0080</device>
>                   </devices>
>             <feature>ALL_FEATURES</feature>     
> <featureStatus>BLOCKED_DEVICE</featureStatus>     
> <driverVersion>8.15.10.2321</driverVersion>     
> <driverVersionComparator>LESS_THAN</driverVersionComparator>   
> </gfxBlacklistEntry>

I can confirm my browser gets this blocklist entry via driver spoof testing against stage.

Comment 19

[Jorge Villalobos [:jorgev] (he/him)]

Benoit, can you confirm that the blocklist in comment #17 looks okay?

Comment 20

[Benoit Jacob [:bjacob] (mostly away)]

I don't see anything shocking here :-) I didn't verify all the hex values.

Comment 21

[Jorge Villalobos [:jorgev] (he/him)]

Then we need someone to figure out why the staged block isn't working...

Comment 22

[Sylvestre Ledru [:Sylvestre]]

Jorge, could you report a bug for this?

@all, if there is no opposition, I would like to land that ASAP. Is there anybody in the European timezone who could do that? thanks

Comment 23

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

I’ve performed some additional investigation today, trying to just see if blocklisting itself should work. See the conclusions and the 6 scenarios in https://docs.google.com/document/d/19UNAXy3KV1-mvoUQt-73pDevdO_3C8DAbvDymefZAM4, which seem to indicate that the solution of pushing the new blocklist XML file will NOT work. This is mainly due to the fact that we do blocklisting in two places: code and XML file, and the two don't mix very well.

Conclusions (issues that would prevent this from working properly):
A. Setting “<os>WINNT</os>” in the blocklist XML may not work as expected, as the code blocklisting just seems to override this (see scenario 3 in the gdoc where the setting has no effect whatsoever).
B. Setting version 8.15.10.2321 in the blocklist XML will not work as internal code (likely in widget/windows/GfxInfo.cpp, but possibly other places as well) just overrides it, and allows D2D and OMTC to be enabled for smaller driver versions (see scenario 5 in the gdoc).
C. Settings in the blocklist XML may not apply unless the user creates a new profile, because of bug 1045055. See scenario 6 in the gdoc, and note that blocklisting from the XML does not update when updating the driver, or when downgrading the driver (similar to updating blocked version in XML). Blocklisting updates correctly for the SAME profile only for settings from the code (widget/windows/GfxInfo.cpp - see bug 1045055).

The other scenarios in the gdoc are meant to show that blocklisting itself works, but only under some conditions. 

Note that the Firefox/browser/blocklist.xml file was used for investigation due to bug 1045055, which requires a new profile to be created in order to verify blocklisting. The new blocklist.xml file is pushed from the server to the profile folder, which basically means this: the new file won't be used when creating a new profile, and blocklisting may not update to block/unblock features when the file itself updates. 

If you can get someone which still has this issue to verify the fix then it could confirm or disprove this, but testing through spoofing indicates that this solution will NOT work.

Comment 24

[Milan Sreckovic [:milan] (needinfo for best results)]

Florin, is this conclusion based on reading the code, or actually running the example through as the users would?  Your conclusion may be correct, but it would surprise me a great deal.

Comment 25

[Jorge Villalobos [:jorgev] (he/him)]

(In reply to Sylvestre Ledru [:sylvestre] from comment #22)
> Jorge, could you report a bug for this?

We could just use this bug for that, but I don't know who is familiar enough with the client side blocklist code to ping about this.

Comment 26

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

(In reply to Milan Sreckovic [:milan] (PTO 10/16-10/17) from comment #24)
> Florin, is this conclusion based on reading the code, or actually running
> the example through as the users would?  Your conclusion may be correct, but
> it would surprise me a great deal.

Milan, conclusion above comes from testing various scenarios by spoofing the video card data, and watching what and how gets blocked. Testing details are documented in the following google doc: https://docs.google.com/document/d/19UNAXy3KV1-mvoUQt-73pDevdO_3C8DAbvDymefZAM4.

Comment 27

[Nicolas Silva [:nical]]

Closing since we'll go for the built-in blacklist.