Both are just tiny one-liners, but as Yoda says, "Size matters not!"
Anyway, I was a touch skeptical at the time that these patches could lead to regressions. Christoph's testing seems to confirm that is indeed the case here.
Of the two, 114 seems the most suspicious since it deals with the memory manager. 115 sounds like it corrects a crash situation, so if it is innocent here it would be nice to keep it.
I have uploaded a fix which disables 114. Please test and verify this resolves the user switching issue. If not, we can also try reverting 115.
The change from 0u1 to 0u2 is the inclusion of these two patches cherrypicked from upstream, to solve bug #344740 which affects i855 chipsets:
114_fix_ xv_with_ non_gem. patch crash_xv_ overlay. patch
115_fix_
Both are just tiny one-liners, but as Yoda says, "Size matters not!"
Anyway, I was a touch skeptical at the time that these patches could lead to regressions. Christoph's testing seems to confirm that is indeed the case here.
Of the two, 114 seems the most suspicious since it deals with the memory manager. 115 sounds like it corrects a crash situation, so if it is innocent here it would be nice to keep it.
I have uploaded a fix which disables 114. Please test and verify this resolves the user switching issue. If not, we can also try reverting 115.