user-level: 0.843
TCG: 0.842
peripherals: 0.804
ppc: 0.795
vnc: 0.771
VMM: 0.769
KVM: 0.765
mistranslation: 0.764
hypervisor: 0.760
register: 0.734
i386: 0.716
x86: 0.712
debug: 0.710
risc-v: 0.704
performance: 0.688
device: 0.674
graphic: 0.667
arm: 0.652
virtual: 0.640
permissions: 0.632
assembly: 0.629
socket: 0.621
architecture: 0.620
PID: 0.616
semantic: 0.612
files: 0.611
kernel: 0.599
boot: 0.579
network: 0.578
--------------------
i386: 0.984
TCG: 0.958
x86: 0.947
assembly: 0.864
debug: 0.129
kernel: 0.060
hypervisor: 0.050
performance: 0.048
architecture: 0.044
files: 0.038
virtual: 0.028
device: 0.021
PID: 0.017
register: 0.015
semantic: 0.015
boot: 0.012
peripherals: 0.011
VMM: 0.006
ppc: 0.005
permissions: 0.005
graphic: 0.004
network: 0.004
user-level: 0.004
risc-v: 0.003
socket: 0.003
arm: 0.002
KVM: 0.002
vnc: 0.001
mistranslation: 0.001

movdqu partial write at page boundary

In TCG mode, when a 16-byte write instruction (such as movdqu) is executed at a page boundary and causes a page fault, a partial write is executed in the first page. See the attached code for an example.

Tested on the qemu-3.0.0-rc1 release.


% gcc -m32 qemu-bug2.c && ./a.out && echo && qemu-i386 ./a.out
*(0x70000ff8+ 0) = aa
*(0x70000ff8+ 1) = aa
*(0x70000ff8+ 2) = aa
*(0x70000ff8+ 3) = aa
*(0x70000ff8+ 4) = aa
*(0x70000ff8+ 5) = aa
*(0x70000ff8+ 6) = aa
*(0x70000ff8+ 7) = aa
*(0x70000ff8+ 8) = 55
*(0x70000ff8+ 9) = 55
*(0x70000ff8+10) = 55
*(0x70000ff8+11) = 55
*(0x70000ff8+12) = 55
*(0x70000ff8+13) = 55
*(0x70000ff8+14) = 55
*(0x70000ff8+15) = 55
page fault: addr=0x70001000 err=0x7
*(0x70000ff8+ 0) = aa
*(0x70000ff8+ 1) = aa
*(0x70000ff8+ 2) = aa
*(0x70000ff8+ 3) = aa
*(0x70000ff8+ 4) = aa
*(0x70000ff8+ 5) = aa
*(0x70000ff8+ 6) = aa
*(0x70000ff8+ 7) = aa
*(0x70000ff8+ 8) = 55
*(0x70000ff8+ 9) = 55
*(0x70000ff8+10) = 55
*(0x70000ff8+11) = 55
*(0x70000ff8+12) = 55
*(0x70000ff8+13) = 55
*(0x70000ff8+14) = 55
*(0x70000ff8+15) = 55

*(0x70000ff8+ 0) = aa
*(0x70000ff8+ 1) = aa
*(0x70000ff8+ 2) = aa
*(0x70000ff8+ 3) = aa
*(0x70000ff8+ 4) = aa
*(0x70000ff8+ 5) = aa
*(0x70000ff8+ 6) = aa
*(0x70000ff8+ 7) = aa
*(0x70000ff8+ 8) = 55
*(0x70000ff8+ 9) = 55
*(0x70000ff8+10) = 55
*(0x70000ff8+11) = 55
*(0x70000ff8+12) = 55
*(0x70000ff8+13) = 55
*(0x70000ff8+14) = 55
*(0x70000ff8+15) = 55
page fault: addr=0x70001000 err=0x6
*(0x70000ff8+ 0) = 77
*(0x70000ff8+ 1) = 66
*(0x70000ff8+ 2) = 55
*(0x70000ff8+ 3) = 44
*(0x70000ff8+ 4) = 33
*(0x70000ff8+ 5) = 22
*(0x70000ff8+ 6) = 11
*(0x70000ff8+ 7) = 0
*(0x70000ff8+ 8) = 55
*(0x70000ff8+ 9) = 55
*(0x70000ff8+10) = 55
*(0x70000ff8+11) = 55
*(0x70000ff8+12) = 55
*(0x70000ff8+13) = 55
*(0x70000ff8+14) = 55
*(0x70000ff8+15) = 55



This is a part of a class of related problems for qemu linux-user, in that any non-atomic store is not validated before initiating a partial write.

For instance, qemu-x86_64, built for arm32, would show this same partial store problem for any 64-bit write crossing a page boundary because we are forced by the limits of the host to split the store into two 32-bit pieces.

While we could probably fix this specific case fairly easily, because it is implemented with an external helper in the first place, we would need some new infrastructure to handle the more general problem.  Exactly what form that should take would need some thought and discussion.

The QEMU project is currently moving its bug tracking to another system.
For this we need to know which bugs are still valid and which could be
closed already. Thus we are setting the bug state to "Incomplete" now.

If the bug has already been fixed in the latest upstream version of QEMU,
then please close this ticket as "Fix released".

If it is not fixed yet and you think that this bug report here is still
valid, then you have two options:

1) If you already have an account on gitlab.com, please open a new ticket
for this problem in our new tracker here:

    https://gitlab.com/qemu-project/qemu/-/issues

and then close this ticket here on Launchpad (or let it expire auto-
matically after 60 days). Please mention the URL of this bug ticket on
Launchpad in the new ticket on GitLab.

2) If you don't have an account on gitlab.com and don't intend to get
one, but still would like to keep this ticket opened, then please switch
the state back to "New" or "Confirmed" within the next 60 days (other-
wise it will get closed as "Expired"). We will then eventually migrate
the ticket automatically to the new system (but you won't be the reporter
of the bug in the new system and thus you won't get notified on changes
anymore).

Thank you and sorry for the inconvenience.


[Expired for QEMU because there has been no activity for 60 days.]