do not sync RBS when changing PT_AR_BSP or PT_CFM
PATCH 2/6 do not sync RBS when changing PT_AR_BSP or PT_CFM
Syncing is no longer needed, because user RBS is already up-to-date. Actually, if a debugger modified the contents of the original RBS prior to changing PT_AR_BSP, the modifications would get overwritten.
Signed-off-by Petr Tesarik <[email protected]>
CC Tony Luck <[email protected]>
CC Christoph Hellwig <[email protected]>
arch/ia64/kernel/ptrace.c +3 -16
@@ -1011,14 +1011,9 @@ # define pt_reg_addr(pt, reg) ((void *)* the kernel was entered. * * Furthermore, when changing the contents of -* PT_AR_BSP (or PT_CFM) we MUST copy any -* users-level stacked registers that are -* stored on the kernel stack back to -* user-space because otherwise, we might end -* up clobbering kernel stacked registers. -* Also, if this happens while the task is -* blocked in a system call, which convert the -* state such that the non-system-call exit +* PT_AR_BSP (or PT_CFM) while the task is +* blocked in a system call, convert the state +* so that the non-system-call exit * path is used. This ensures that the proper * state will be picked up when resuming * execution. However, it *also* means that
@@ -1035,10 +1030,6 @@ # define pt_reg_addr(pt, reg) ((void *) urbs_end =ia64_get_user_rbs_end (child, pt, &cfm);if (write_access) {if (*data != urbs_end) { -if (ia64_sync_user_rbs (child, sw, - pt->ar_bspstore, - urbs_end) <0 ) -return -1 ;if (in_syscall (pt)) convert_to_non_syscall(child, pt,
@@ -1058,10 +1049,6 @@ # define pt_reg_addr(pt, reg) ((void *) urbs_end =ia64_get_user_rbs_end (child, pt, &cfm);if (write_access) {if (((cfm ^ *data) &PFM_MASK ) !=0 ) { -if (ia64_sync_user_rbs (child, sw, - pt->ar_bspstore, - urbs_end) <0 ) -return -1 ;if (in_syscall (pt)) convert_to_non_syscall(child, pt,