It is the same set of patches as we have on Bionic.
Bionic has
1. 8fea70440eb0d095442de7e80d586a285cf96be5
2. 399b2896d4948a1ec0278d896ea3a561df768d64
3. 8c5909c41916f25b47bfdc465059a926603c1319
4. 8ff43ee404d3e295839d1fd4e9e6571ca7a62a66
Disco for this bug has #2+#4 while #1+#3 are already part of the base version that is in qemu of Disco.
Due to different contexts they are slightly different.
Upstream defines it as
+#define SPAPR_CAP_CCF_ASSIST 0x09
Due to the context change in Bionic and Disco 0x06 and 0x08 respectively.
That index matters if it would be off in the capability_table[SPAPR_CAP_NUM].
I recounted the field to ensure there is no off by one and also otherwise compared the diffs of the upstream commits and the bionic/disco backports. There doesn't seem to be an issue in those.
@Michael could you retest this on Disco and the kernel you used (and worked) from Bionic.
If it is a kernel issue I'm fine and we can open a kernel task for it for Disco? That would help as we would not have to stop/gate qemu in that case.
It is the same set of patches as we have on Bionic. 5442de7e80d586a 285cf96be5 ec0278d896ea3a5 61df768d64 b47bfdc465059a9 26603c1319 5839d1fd4e9e657 1ca7a62a66
Bionic has
1. 8fea70440eb0d09
2. 399b2896d4948a1
3. 8c5909c41916f25
4. 8ff43ee404d3e29
Disco for this bug has #2+#4 while #1+#3 are already part of the base version that is in qemu of Disco.
Due to different contexts they are slightly different. CCF_ASSIST 0x09 table[SPAPR_ CAP_NUM] .
Upstream defines it as
+#define SPAPR_CAP_
Due to the context change in Bionic and Disco 0x06 and 0x08 respectively.
That index matters if it would be off in the capability_
I recounted the field to ensure there is no off by one and also otherwise compared the diffs of the upstream commits and the bionic/disco backports. There doesn't seem to be an issue in those.
@Michael could you retest this on Disco and the kernel you used (and worked) from Bionic.
If it is a kernel issue I'm fine and we can open a kernel task for it for Disco? That would help as we would not have to stop/gate qemu in that case.