I think RTM does a RP and RP is setup for a 16 byte PSW you think they could of found 8 bytes in SDWARC4
> On Jun 26, 2019, at 3:42 PM, Jonathan Scott <jonathan_sc...@vnet.ibm.com> > wrote: > > Ref: Your note of Wed, 26 Jun 2019 09:43:30 -0400 > > If you have the linkage stack available and storage containing > the information needed to resume (such as an SDWA or copy) then > I think it is possible to create code to resume (at least in > problem state) at a chosen address with all registers containing > their original values, original PSW ASC mode, AMODE, condition > code and program mask, using BAKR and PR. > > I originally wrote some code like this in the 1990s to implement > a sort of problem-state "must complete" where the original code > could be resumed for a few instructions from the recovery > routine, and when it turned off "must complete" it would return > control to the recovery routine. However, it was never used in > production, and I'm not sure whether it was even tested, so > this method comes with no guarantees. > > Basically, you set up the registers that are returned by PR, put > the resume address in a register that isn't returned by PR, > issue BAKR to save the resume address, restore the other > registers and issue PR to resume. However, if you want to do it > properly with condition code, ASC mode and so on, it's quite > tricky. > > - Start in primary mode with R1 pointing to the SDWA. > > - Load access registers 2 through 14 (using one of the > other registers to address that SDWA section). > > - Load general registers R2 through R14. > > - Get the program mask (with CC) and ASC mode in R15. > > - Get the PSW address and AMODE indicator bits in R0 (because R1 > and R15 are still needed). > > - Use SAC to set the ASC mode and SPM to set program mask. > > - Copy the PSW instruction address and AMODE from R0 to R15. > > - Issue BAKR R15,0 to stack the PSW and registers R2 to R14. > > - Switch back to primary mode to access SDWA again. > > - Get the program mask again in R15. > > - Repeat the SPM in case PR doesn't restore the condition code > (it is undefined whether PR restores the condition code, but > it seems reasonable to assume it either takes the stacked > one or preserves the current one). > > - Load access registers 15 through 1 from the SDWA, using some > other register to address the relevant part. > > - Load general registers R15 through R1. > > - Issue PR to resume execution with original state restored. > > I hope I've remembered that correctly! I probably have some > code around somewhere to do that, but I can't remember where. > > Jonathan Scott, HLASM > IBM Hursley, UK
