fridtjof/qemu-cr16
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ppc/xive2: Support redistribution of group interrupts

Browse source 
When an XIVE context is pulled while it has an active, unacknowledged
group interrupt, XIVE will check to see if a context on another thread
can handle the interrupt and, if so, notify that context.  If there
are no contexts that can handle the interrupt, then the interrupt is
added to a backlog and XIVE will attempt to escalate the interrupt,
if configured to do so, allowing the higher privileged handler to
activate a context that can handle the original interrupt.

Signed-off-by: Glenn Miles <milesg@linux.ibm.com>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Michael Kowal <kowal@linux.ibm.com>
Tested-by: Gautam Menghani <gautam@linux.ibm.com>
Link: https://lore.kernel.org/qemu-devel/20250512031100.439842-23-npiggin@gmail.com
Signed-off-by: Cédric Le Goater <clg@redhat.com>
This commit is contained in:
Glenn Miles 2025-05-12 13:10:31 +10:00 • committed by Cédric Le Goater
parent b22ffb4299
commit 555e446019
2 changed files with 83 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

84
hw/intc/xive2.c
View file

@ -19,6 +19,10 @@
#include "hw/ppc/xive2_regs.h"
#include "trace.h"
static void xive2_router_end_notify(Xive2Router *xrtr, uint8_t end_blk,
uint32_t end_idx, uint32_t end_data,
bool redistribute);
uint32_t xive2_router_get_config(Xive2Router *xrtr)
{
Xive2RouterClass *xrc = XIVE2_ROUTER_GET_CLASS(xrtr);
@ -597,6 +601,68 @@ static uint32_t xive2_tctx_hw_cam_line(XivePresenter *xptr, XiveTCTX *tctx)
return xive2_nvp_cam_line(blk, 1 << tid_shift | (pir & tid_mask));
}
static void xive2_redistribute(Xive2Router *xrtr, XiveTCTX *tctx,
uint8_t nvp_blk, uint32_t nvp_idx, uint8_t ring)
{
uint8_t nsr = tctx->regs[ring + TM_NSR];
uint8_t crowd = NVx_CROWD_LVL(nsr);
uint8_t group = NVx_GROUP_LVL(nsr);
uint8_t nvgc_blk;
uint8_t nvgc_idx;
uint8_t end_blk;
uint32_t end_idx;
uint8_t pipr = tctx->regs[ring + TM_PIPR];
Xive2Nvgc nvgc;
uint8_t prio_limit;
uint32_t cfg;
/* convert crowd/group to blk/idx */
if (group > 0) {
nvgc_idx = (nvp_idx & (0xffffffff << group)) |
((1 << (group - 1)) - 1);
} else {
nvgc_idx = nvp_idx;
}
if (crowd > 0) {
crowd = (crowd == 3) ? 4 : crowd;
nvgc_blk = (nvp_blk & (0xffffffff << crowd)) |
((1 << (crowd - 1)) - 1);
} else {
nvgc_blk = nvp_blk;
}
/* Use blk/idx to retrieve the NVGC */
if (xive2_router_get_nvgc(xrtr, crowd, nvgc_blk, nvgc_idx, &nvgc)) {
qemu_log_mask(LOG_GUEST_ERROR, "XIVE: no %s %x/%x\n",
crowd ? "NVC" : "NVG", nvgc_blk, nvgc_idx);
return;
}
/* retrieve the END blk/idx from the NVGC */
end_blk = xive_get_field32(NVGC2_W1_END_BLK, nvgc.w1);
end_idx = xive_get_field32(NVGC2_W1_END_IDX, nvgc.w1);
/* determine number of priorities being used */
cfg = xive2_router_get_config(xrtr);
if (cfg & XIVE2_EN_VP_GRP_PRIORITY) {
prio_limit = 1 << GETFIELD(NVGC2_W1_PSIZE, nvgc.w1);
} else {
prio_limit = 1 << GETFIELD(XIVE2_VP_INT_PRIO, cfg);
}
/* add priority offset to end index */
end_idx += pipr % prio_limit;
/* trigger the group END */
xive2_router_end_notify(xrtr, end_blk, end_idx, 0, true);
/* clear interrupt indication for the context */
tctx->regs[ring + TM_NSR] = 0;
tctx->regs[ring + TM_PIPR] = tctx->regs[ring + TM_CPPR];
xive_tctx_reset_signal(tctx, ring);
}
static uint64_t xive2_tm_pull_ctx(XivePresenter *xptr, XiveTCTX *tctx,
hwaddr offset, unsigned size, uint8_t ring)
{
@ -608,6 +674,7 @@ static uint64_t xive2_tm_pull_ctx(XivePresenter *xptr, XiveTCTX *tctx,
uint8_t cur_ring;
bool valid;
bool do_save;
uint8_t nsr;
xive2_cam_decode(cam, &nvp_blk, &nvp_idx, &valid, &do_save);
@ -624,6 +691,12 @@ static uint64_t xive2_tm_pull_ctx(XivePresenter *xptr, XiveTCTX *tctx,
memcpy(&tctx->regs[cur_ring + TM_WORD2], &ringw2_new, 4);
}
/* Active group/crowd interrupts need to be redistributed */
nsr = tctx->regs[ring + TM_NSR];
if (xive_nsr_indicates_group_exception(ring, nsr)) {
xive2_redistribute(xrtr, tctx, nvp_blk, nvp_idx, ring);
}
if (xive2_router_get_config(xrtr) & XIVE2_VP_SAVE_RESTORE && do_save) {
xive2_tctx_save_ctx(xrtr, tctx, nvp_blk, nvp_idx, ring);
}
@ -1352,7 +1425,8 @@ static bool xive2_router_end_es_notify(Xive2Router *xrtr, uint8_t end_blk,
* message has the same parameters than in the function below.
*/
static void xive2_router_end_notify(Xive2Router *xrtr, uint8_t end_blk,
uint32_t end_idx, uint32_t end_data)
uint32_t end_idx, uint32_t end_data,
bool redistribute)
{
Xive2End end;
uint8_t priority;
@ -1380,7 +1454,7 @@ static void xive2_router_end_notify(Xive2Router *xrtr, uint8_t end_blk,
return;
}
if (xive2_end_is_enqueue(&end)) {
if (!redistribute && xive2_end_is_enqueue(&end)) {
xive2_end_enqueue(&end, end_data);
/* Enqueuing event data modifies the EQ toggle and index */
xive2_router_write_end(xrtr, end_blk, end_idx, &end, 1);
@ -1560,7 +1634,8 @@ do_escalation:
xive2_router_end_notify(xrtr,
xive_get_field32(END2_W4_END_BLOCK, end.w4),
xive_get_field32(END2_W4_ESC_END_INDEX, end.w4),
xive_get_field32(END2_W5_ESC_END_DATA, end.w5));
xive_get_field32(END2_W5_ESC_END_DATA, end.w5),
false);
} /* end END adaptive escalation */
else {
@ -1641,7 +1716,8 @@ void xive2_notify(Xive2Router *xrtr , uint32_t lisn, bool pq_checked)
xive2_router_end_notify(xrtr,
xive_get_field64(EAS2_END_BLOCK, eas.w),
xive_get_field64(EAS2_END_INDEX, eas.w),
xive_get_field64(EAS2_END_DATA, eas.w));
xive_get_field64(EAS2_END_DATA, eas.w),
false);
return;
}

3
include/hw/ppc/xive2_regs.h
View file

@ -224,6 +224,9 @@ typedef struct Xive2Nvgc {
#define NVGC2_W0_VALID PPC_BIT32(0)
#define NVGC2_W0_PGONEXT PPC_BITMASK32(26, 31)
uint32_t w1;
#define NVGC2_W1_PSIZE PPC_BITMASK32(0, 1)
#define NVGC2_W1_END_BLK PPC_BITMASK32(4, 7)
#define NVGC2_W1_END_IDX PPC_BITMASK32(8, 31)
uint32_t w2;
uint32_t w3;
uint32_t w4;