xiaomi-sm8450/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

sparc64: Store per-cpu offset in trap_block[]

Browse Source 
Surprisingly this actually makes LOAD_PER_CPU_BASE() a little
more efficient.

Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
David S. Miller
2009-04-01 01:47:10 -07:00
parent 19f0fa3fb3
commit 5a5488d3bb
5 changed files with 21 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
arch/sparc/kernel/head_64.S
View File

@@ -641,28 +641,6 @@ tlb_fixup_done:
/* Not reached... */
1:
/* If we boot on a non-zero cpu, all of the per-cpu
* variable references we make before setting up the
* per-cpu areas will use a bogus offset. Put a
* compensating factor into __per_cpu_base to handle
* this cleanly.
*
* What the per-cpu code calculates is:
*
* __per_cpu_base + (cpu << __per_cpu_shift)
*
* These two variables are zero initially, so to
* make it all cancel out to zero we need to put
* "0 - (cpu << 0)" into __per_cpu_base so that the
* above formula evaluates to zero.
*
* We cannot even perform a printk() until this stuff
* is setup as that calls cpu_clock() which uses
* per-cpu variables.
*/
sub %g0, %o0, %o1
sethi %hi(__per_cpu_base), %o2
stx %o1, [%o2 + %lo(__per_cpu_base)]
#else
mov 0, %o0
#endif