From: Sebastian Ott <sebott@linux.ibm.com>
Date: Thu, 27 Sep 2018 13:57:12 +0200
Subject: s390/pci: provide support for CPU directed interrupts
Git-commit: e979ce7bced2ee019b5b1a040295484bd7f23680
Patch-mainline: v5.2-rc1
References: jsc#SLE-5789 FATE#327042 bsc#1134730 LTC#173388
Up until now all interrupts on s390 have been floating. For MSI interrupts
we've used a global summary bit vector (with a bit for each function) and
a per-function interrupt bit vector (with a bit per MSI).
This patch introduces a new IRQ delivery mode: CPU directed interrupts.
In this new mode a per-CPU interrupt bit vector is used (with a bit per
MSI per function). Further it is now possible to direct an IRQ to a
specific CPU so we can finally support IRQ affinity.
If an interrupt can't be delivered because the appointed CPU is occupied
by a hypervisor the interrupt is delivered floating. For this a global
summary bit vector is used (with a bit per CPU).
Signed-off-by: Sebastian Ott <sebott@linux.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Acked-by: Petr Tesarik <ptesarik@suse.com>
---
arch/s390/include/asm/pci.h | 2
arch/s390/include/asm/pci_clp.h | 6
arch/s390/include/asm/pci_insn.h | 74 +++++++-
arch/s390/pci/pci_insn.c | 10 -
arch/s390/pci/pci_irq.c | 339 +++++++++++++++++++++++++++++++++------
5 files changed, 368 insertions(+), 63 deletions(-)
--- a/arch/s390/include/asm/pci.h
+++ b/arch/s390/include/asm/pci.h
@@ -110,6 +110,8 @@ struct zpci_dev {
/* IRQ stuff */
u64 msi_addr; /* MSI address */
unsigned int max_msi; /* maximum number of MSI's */
+ unsigned int msi_first_bit;
+ unsigned int msi_nr_irqs;
struct airq_iv *aibv; /* adapter interrupt bit vector */
unsigned long aisb; /* number of the summary bit */
--- a/arch/s390/include/asm/pci_clp.h
+++ b/arch/s390/include/asm/pci_clp.h
@@ -117,7 +117,11 @@ struct clp_rsp_query_pci_grp {
u8 refresh : 1; /* TLB refresh mode */
u16 reserved2;
u16 mui;
- u64 reserved3;
+ u16 : 16;
+ u16 maxfaal;
+ u16 : 4;
+ u16 dnoi : 12;
+ u16 maxcpu;
u64 dasm; /* dma address space mask */
u64 msia; /* MSI address */
u64 reserved4;
--- a/arch/s390/include/asm/pci_insn.h
+++ b/arch/s390/include/asm/pci_insn.h
@@ -37,6 +37,8 @@
#define ZPCI_MOD_FC_RESET_ERROR 7
#define ZPCI_MOD_FC_RESET_BLOCK 9
#define ZPCI_MOD_FC_SET_MEASURE 10
+#define ZPCI_MOD_FC_REG_INT_D 16
+#define ZPCI_MOD_FC_DEREG_INT_D 17
/* FIB function controls */
#define ZPCI_FIB_FC_ENABLED 0x80
@@ -50,16 +52,7 @@
#define ZPCI_FIB_FC_LS_BLOCKED 0x20
#define ZPCI_FIB_FC_DMAAS_REG 0x10
-/* Function Information Block */
-struct zpci_fib {
- u32 fmt : 8; /* format */
- u32 : 24;
- u32 : 32;
- u8 fc; /* function controls */
- u64 : 56;
- u64 pba; /* PCI base address */
- u64 pal; /* PCI address limit */
- u64 iota; /* I/O Translation Anchor */
+struct zpci_fib_fmt0 {
u32 : 1;
u32 isc : 3; /* Interrupt subclass */
u32 noi : 12; /* Number of interrupts */
@@ -71,16 +64,75 @@ struct zpci_fib {
u32 : 32;
u64 aibv; /* Adapter int bit vector address */
u64 aisb; /* Adapter int summary bit address */
+};
+
+struct zpci_fib_fmt1 {
+ u32 : 4;
+ u32 noi : 12;
+ u32 : 16;
+ u32 dibvo : 16;
+ u32 : 16;
+ u64 : 64;
+ u64 : 64;
+};
+
+/* Function Information Block */
+struct zpci_fib {
+ u32 fmt : 8; /* format */
+ u32 : 24;
+ u32 : 32;
+ u8 fc; /* function controls */
+ u64 : 56;
+ u64 pba; /* PCI base address */
+ u64 pal; /* PCI address limit */
+ u64 iota; /* I/O Translation Anchor */
+ union {
+ struct zpci_fib_fmt0 fmt0;
+ struct zpci_fib_fmt1 fmt1;
+ };
u64 fmb_addr; /* Function measurement block address and key */
u32 : 32;
u32 gd;
} __packed __aligned(8);
+/* directed interruption information block */
+struct zpci_diib {
+ u32 : 1;
+ u32 isc : 3;
+ u32 : 28;
+ u16 : 16;
+ u16 nr_cpus;
+ u64 disb_addr;
+ u64 : 64;
+ u64 : 64;
+} __packed __aligned(8);
+
+/* cpu directed interruption information block */
+struct zpci_cdiib {
+ u64 : 64;
+ u64 dibv_addr;
+ u64 : 64;
+ u64 : 64;
+ u64 : 64;
+} __packed __aligned(8);
+
+union zpci_sic_iib {
+ struct zpci_diib diib;
+ struct zpci_cdiib cdiib;
+};
+
u8 zpci_mod_fc(u64 req, struct zpci_fib *fib, u8 *status);
int zpci_refresh_trans(u64 fn, u64 addr, u64 range);
int zpci_load(u64 *data, u64 req, u64 offset);
int zpci_store(u64 data, u64 req, u64 offset);
int zpci_store_block(const u64 *data, u64 req, u64 offset);
-int zpci_set_irq_ctrl(u16 ctl, char *unused, u8 isc);
+int __zpci_set_irq_ctrl(u16 ctl, u8 isc, union zpci_sic_iib *iib);
+
+static inline int zpci_set_irq_ctrl(u16 ctl, u8 isc)
+{
+ union zpci_sic_iib iib = {{0}};
+
+ return __zpci_set_irq_ctrl(ctl, isc, &iib);
+}
#endif
--- a/arch/s390/pci/pci_insn.c
+++ b/arch/s390/pci/pci_insn.c
@@ -95,13 +95,15 @@ int zpci_refresh_trans(u64 fn, u64 addr,
}
/* Set Interruption Controls */
-int zpci_set_irq_ctrl(u16 ctl, char *unused, u8 isc)
+int __zpci_set_irq_ctrl(u16 ctl, u8 isc, union zpci_sic_iib *iib)
{
if (!test_facility(72))
return -EIO;
- asm volatile (
- " .insn rsy,0xeb00000000d1,%[ctl],%[isc],%[u]\n"
- : : [ctl] "d" (ctl), [isc] "d" (isc << 27), [u] "Q" (*unused));
+
+ asm volatile(
+ ".insn rsy,0xeb00000000d1,%[ctl],%[isc],%[iib]\n"
+ : : [ctl] "d" (ctl), [isc] "d" (isc << 27), [iib] "Q" (*iib));
+
return 0;
}
--- a/arch/s390/pci/pci_irq.c
+++ b/arch/s390/pci/pci_irq.c
@@ -7,20 +7,31 @@
#include <linux/kernel_stat.h>
#include <linux/pci.h>
#include <linux/msi.h>
+#include <linux/smp.h>
#include <asm/isc.h>
#include <asm/airq.h>
+static enum {FLOATING, DIRECTED} irq_delivery;
+
#define SIC_IRQ_MODE_ALL 0
#define SIC_IRQ_MODE_SINGLE 1
+#define SIC_IRQ_MODE_DIRECT 4
+#define SIC_IRQ_MODE_D_ALL 16
+#define SIC_IRQ_MODE_D_SINGLE 17
+#define SIC_IRQ_MODE_SET_CPU 18
/*
- * summary bit vector - one summary bit per function
+ * summary bit vector
+ * FLOATING - summary bit per function
+ * DIRECTED - summary bit per cpu (only used in fallback path)
*/
static struct airq_iv *zpci_sbv;
/*
- * interrupt bit vectors - one vector per function
+ * interrupt bit vectors
+ * FLOATING - interrupt bit vector per function
+ * DIRECTED - interrupt bit vector per cpu
*/
static struct airq_iv **zpci_ibv;
@@ -31,13 +42,13 @@ static int zpci_set_airq(struct zpci_dev
struct zpci_fib fib = {0};
u8 status;
- fib.isc = PCI_ISC;
- fib.sum = 1; /* enable summary notifications */
- fib.noi = airq_iv_end(zdev->aibv);
- fib.aibv = (unsigned long) zdev->aibv->vector;
- fib.aibvo = 0; /* each zdev has its own interrupt vector */
- fib.aisb = (unsigned long) zpci_sbv->vector + (zdev->aisb/64)*8;
- fib.aisbo = zdev->aisb & 63;
+ fib.fmt0.isc = PCI_ISC;
+ fib.fmt0.sum = 1; /* enable summary notifications */
+ fib.fmt0.noi = airq_iv_end(zdev->aibv);
+ fib.fmt0.aibv = (unsigned long) zdev->aibv->vector;
+ fib.fmt0.aibvo = 0; /* each zdev has its own interrupt vector */
+ fib.fmt0.aisb = (unsigned long) zpci_sbv->vector + (zdev->aisb/64)*8;
+ fib.fmt0.aisbo = zdev->aisb & 63;
return zpci_mod_fc(req, &fib, &status) ? -EIO : 0;
}
@@ -57,13 +68,134 @@ static int zpci_clear_airq(struct zpci_d
return cc ? -EIO : 0;
}
+/* Modify PCI: Register CPU directed interruptions */
+static int zpci_set_directed_irq(struct zpci_dev *zdev)
+{
+ u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_REG_INT_D);
+ struct zpci_fib fib = {0};
+ u8 status;
+
+ fib.fmt = 1;
+ fib.fmt1.noi = zdev->msi_nr_irqs;
+ fib.fmt1.dibvo = zdev->msi_first_bit;
+
+ return zpci_mod_fc(req, &fib, &status) ? -EIO : 0;
+}
+
+/* Modify PCI: Unregister CPU directed interruptions */
+static int zpci_clear_directed_irq(struct zpci_dev *zdev)
+{
+ u64 req = ZPCI_CREATE_REQ(zdev->fh, 0, ZPCI_MOD_FC_DEREG_INT_D);
+ struct zpci_fib fib = {0};
+ u8 cc, status;
+
+ fib.fmt = 1;
+ cc = zpci_mod_fc(req, &fib, &status);
+ if (cc == 3 || (cc == 1 && status == 24))
+ /* Function already gone or IRQs already deregistered. */
+ cc = 0;
+
+ return cc ? -EIO : 0;
+}
+
+static int zpci_set_irq_affinity(struct irq_data *data, const struct cpumask *dest,
+ bool force)
+{
+ struct msi_desc *entry = irq_get_msi_desc(data->irq);
+ struct msi_msg msg = entry->msg;
+
+ msg.address_lo &= 0xff0000ff;
+ msg.address_lo |= (cpumask_first(dest) << 8);
+ pci_write_msi_msg(data->irq, &msg);
+
+ return IRQ_SET_MASK_OK;
+}
+
static struct irq_chip zpci_irq_chip = {
.name = "zPCI",
.irq_unmask = pci_msi_unmask_irq,
.irq_mask = pci_msi_mask_irq,
+ .irq_set_affinity = zpci_set_irq_affinity,
+};
+
+static void zpci_handle_cpu_local_irq(bool rescan)
+{
+ struct airq_iv *dibv = zpci_ibv[smp_processor_id()];
+ unsigned long bit;
+ int irqs_on = 0;
+
+ for (bit = 0;;) {
+ /* Scan the directed IRQ bit vector */
+ bit = airq_iv_scan(dibv, bit, airq_iv_end(dibv));
+ if (bit == -1UL) {
+ if (!rescan || irqs_on++)
+ /* End of second scan with interrupts on. */
+ break;
+ /* First scan complete, reenable interrupts. */
+ if (zpci_set_irq_ctrl(SIC_IRQ_MODE_D_SINGLE, PCI_ISC))
+ break;
+ bit = 0;
+ continue;
+ }
+ inc_irq_stat(IRQIO_MSI);
+ generic_handle_irq(airq_iv_get_data(dibv, bit));
+ }
+}
+
+struct cpu_irq_data {
+ call_single_data_t csd;
+ atomic_t scheduled;
};
+static DEFINE_PER_CPU_SHARED_ALIGNED(struct cpu_irq_data, irq_data);
+
+static void zpci_handle_remote_irq(void *data)
+{
+ atomic_t *scheduled = data;
+
+ do {
+ zpci_handle_cpu_local_irq(false);
+ } while (atomic_dec_return(scheduled));
+}
+
+static void zpci_handle_fallback_irq(void)
+{
+ struct cpu_irq_data *cpu_data;
+ unsigned long cpu;
+ int irqs_on = 0;
+
+ for (cpu = 0;;) {
+ cpu = airq_iv_scan(zpci_sbv, cpu, airq_iv_end(zpci_sbv));
+ if (cpu == -1UL) {
+ if (irqs_on++)
+ /* End of second scan with interrupts on. */
+ break;
+ /* First scan complete, reenable interrupts. */
+ if (zpci_set_irq_ctrl(SIC_IRQ_MODE_SINGLE, PCI_ISC))
+ break;
+ cpu = 0;
+ continue;
+ }
+ cpu_data = &per_cpu(irq_data, cpu);
+ if (atomic_inc_return(&cpu_data->scheduled) > 1)
+ continue;
+
+ cpu_data->csd.func = zpci_handle_remote_irq;
+ cpu_data->csd.info = &cpu_data->scheduled;
+ cpu_data->csd.flags = 0;
+ smp_call_function_single_async(cpu, &cpu_data->csd);
+ }
+}
+
+static void zpci_directed_irq_handler(struct airq_struct *airq, bool floating)
+{
+ inc_irq_stat(IRQIO_PCI);
+ if (floating)
+ zpci_handle_fallback_irq();
+ else
+ zpci_handle_cpu_local_irq(true);
+}
-static void zpci_irq_handler(struct airq_struct *airq, bool floating)
+static void zpci_floating_irq_handler(struct airq_struct *airq, bool floating)
{
unsigned long si, ai;
struct airq_iv *aibv;
@@ -78,7 +210,7 @@ static void zpci_irq_handler(struct airq
/* End of second scan with interrupts on. */
break;
/* First scan complete, reenable interrupts. */
- if (zpci_set_irq_ctrl(SIC_IRQ_MODE_SINGLE, NULL, PCI_ISC))
+ if (zpci_set_irq_ctrl(SIC_IRQ_MODE_SINGLE, PCI_ISC))
break;
si = 0;
continue;
@@ -101,54 +233,79 @@ static void zpci_irq_handler(struct airq
int arch_setup_msi_irqs(struct pci_dev *pdev, int nvec, int type)
{
struct zpci_dev *zdev = to_zpci(pdev);
- unsigned int hwirq, msi_vecs;
- unsigned long aisb;
+ unsigned int hwirq, msi_vecs, cpu;
+ unsigned long bit;
struct msi_desc *msi;
struct msi_msg msg;
int rc, irq;
zdev->aisb = -1UL;
+ zdev->msi_first_bit = -1U;
if (type == PCI_CAP_ID_MSI && nvec > 1)
return 1;
msi_vecs = min_t(unsigned int, nvec, zdev->max_msi);
- /* Allocate adapter summary indicator bit */
- aisb = airq_iv_alloc_bit(zpci_sbv);
- if (aisb == -1UL)
- return -EIO;
- zdev->aisb = aisb;
-
- /* Create adapter interrupt vector */
- zdev->aibv = airq_iv_create(msi_vecs, AIRQ_IV_DATA | AIRQ_IV_BITLOCK);
- if (!zdev->aibv)
- return -ENOMEM;
+ if (irq_delivery == DIRECTED) {
+ /* Allocate cpu vector bits */
+ bit = airq_iv_alloc(zpci_ibv[0], msi_vecs);
+ if (bit == -1UL)
+ return -EIO;
+ } else {
+ /* Allocate adapter summary indicator bit */
+ bit = airq_iv_alloc_bit(zpci_sbv);
+ if (bit == -1UL)
+ return -EIO;
+ zdev->aisb = bit;
+
+ /* Create adapter interrupt vector */
+ zdev->aibv = airq_iv_create(msi_vecs, AIRQ_IV_DATA | AIRQ_IV_BITLOCK);
+ if (!zdev->aibv)
+ return -ENOMEM;
- /* Wire up shortcut pointer */
- zpci_ibv[aisb] = zdev->aibv;
+ /* Wire up shortcut pointer */
+ zpci_ibv[bit] = zdev->aibv;
+ /* Each function has its own interrupt vector */
+ bit = 0;
+ }
/* Request MSI interrupts */
- hwirq = 0;
+ hwirq = bit;
for_each_pci_msi_entry(msi, pdev) {
- if (hwirq >= msi_vecs)
+ rc = -EIO;
+ if (hwirq - bit >= msi_vecs)
break;
- irq = irq_alloc_desc(0); /* Alloc irq on node 0 */
+ irq = __irq_alloc_descs(-1, 0, 1, 0, THIS_MODULE, msi->affinity);
if (irq < 0)
return -ENOMEM;
rc = irq_set_msi_desc(irq, msi);
if (rc)
return rc;
irq_set_chip_and_handler(irq, &zpci_irq_chip,
- handle_simple_irq);
+ handle_percpu_irq);
msg.data = hwirq;
- msg.address_lo = zdev->msi_addr & 0xffffffff;
+ if (irq_delivery == DIRECTED) {
+ msg.address_lo = zdev->msi_addr & 0xff0000ff;
+ msg.address_lo |= msi->affinity ?
+ (cpumask_first(msi->affinity) << 8) : 0;
+ for_each_possible_cpu(cpu) {
+ airq_iv_set_data(zpci_ibv[cpu], hwirq, irq);
+ }
+ } else {
+ msg.address_lo = zdev->msi_addr & 0xffffffff;
+ airq_iv_set_data(zdev->aibv, hwirq, irq);
+ }
msg.address_hi = zdev->msi_addr >> 32;
pci_write_msi_msg(irq, &msg);
- airq_iv_set_data(zdev->aibv, hwirq, irq);
hwirq++;
}
- /* Enable adapter interrupts */
- rc = zpci_set_airq(zdev);
+ zdev->msi_first_bit = bit;
+ zdev->msi_nr_irqs = msi_vecs;
+
+ if (irq_delivery == DIRECTED)
+ rc = zpci_set_directed_irq(zdev);
+ else
+ rc = zpci_set_airq(zdev);
if (rc)
return rc;
@@ -161,8 +318,11 @@ void arch_teardown_msi_irqs(struct pci_d
struct msi_desc *msi;
int rc;
- /* Disable adapter interrupts */
- rc = zpci_clear_airq(zdev);
+ /* Disable interrupts */
+ if (irq_delivery == DIRECTED)
+ rc = zpci_clear_directed_irq(zdev);
+ else
+ rc = zpci_clear_airq(zdev);
if (rc)
return;
@@ -191,37 +351,114 @@ void arch_teardown_msi_irqs(struct pci_d
airq_iv_release(zdev->aibv);
zdev->aibv = NULL;
}
+
+ if ((irq_delivery == DIRECTED) && zdev->msi_first_bit != -1U)
+ airq_iv_free(zpci_ibv[0], zdev->msi_first_bit, zdev->msi_nr_irqs);
}
static struct airq_struct zpci_airq = {
- .handler = zpci_irq_handler,
+ .handler = zpci_floating_irq_handler,
.isc = PCI_ISC,
};
-int __init zpci_irq_init(void)
+static void __init cpu_enable_directed_irq(void *unused)
{
- int rc;
+ union zpci_sic_iib iib = {{0}};
- rc = register_adapter_interrupt(&zpci_airq);
- if (rc)
- goto out;
- /* Set summary to 1 to be called every time for the ISC. */
- *zpci_airq.lsi_ptr = 1;
+ iib.cdiib.dibv_addr = (u64) zpci_ibv[smp_processor_id()]->vector;
+
+ __zpci_set_irq_ctrl(SIC_IRQ_MODE_SET_CPU, 0, &iib);
+ zpci_set_irq_ctrl(SIC_IRQ_MODE_D_SINGLE, PCI_ISC);
+}
+
+static int __init zpci_directed_irq_init(void)
+{
+ union zpci_sic_iib iib = {{0}};
+ unsigned int cpu;
+
+ zpci_sbv = airq_iv_create(num_possible_cpus(), 0);
+ if (!zpci_sbv)
+ return -ENOMEM;
+
+ iib.diib.isc = PCI_ISC;
+ iib.diib.nr_cpus = num_possible_cpus();
+ iib.diib.disb_addr = (u64) zpci_sbv->vector;
+ __zpci_set_irq_ctrl(SIC_IRQ_MODE_DIRECT, 0, &iib);
+
+ zpci_ibv = kcalloc(num_possible_cpus(), sizeof(*zpci_ibv),
+ GFP_KERNEL);
+ if (!zpci_ibv)
+ return -ENOMEM;
+
+ for_each_possible_cpu(cpu) {
+ /*
+ * Per CPU IRQ vectors look the same but bit-allocation
+ * is only done on the first vector.
+ */
+ zpci_ibv[cpu] = airq_iv_create(cache_line_size() * BITS_PER_BYTE,
+ AIRQ_IV_DATA |
+ AIRQ_IV_CACHELINE |
+ (!cpu ? AIRQ_IV_ALLOC : 0));
+ if (!zpci_ibv[cpu])
+ return -ENOMEM;
+ }
+ on_each_cpu(cpu_enable_directed_irq, NULL, 1);
- rc = -ENOMEM;
+ zpci_irq_chip.irq_set_affinity = zpci_set_irq_affinity;
+
+ return 0;
+}
+
+static int __init zpci_floating_irq_init(void)
+{
zpci_ibv = kcalloc(ZPCI_NR_DEVICES, sizeof(*zpci_ibv), GFP_KERNEL);
if (!zpci_ibv)
- goto out_airq;
+ return -ENOMEM;
zpci_sbv = airq_iv_create(ZPCI_NR_DEVICES, AIRQ_IV_ALLOC);
if (!zpci_sbv)
goto out_free;
- zpci_set_irq_ctrl(SIC_IRQ_MODE_SINGLE, NULL, PCI_ISC);
return 0;
out_free:
kfree(zpci_ibv);
+ return -ENOMEM;
+}
+
+int __init zpci_irq_init(void)
+{
+ int rc;
+
+ irq_delivery = sclp.has_dirq ? DIRECTED : FLOATING;
+ if (irq_delivery == DIRECTED)
+ zpci_airq.handler = zpci_directed_irq_handler;
+
+ rc = register_adapter_interrupt(&zpci_airq);
+ if (rc)
+ goto out;
+ /* Set summary to 1 to be called every time for the ISC. */
+ *zpci_airq.lsi_ptr = 1;
+
+ switch (irq_delivery) {
+ case FLOATING:
+ rc = zpci_floating_irq_init();
+ break;
+ case DIRECTED:
+ rc = zpci_directed_irq_init();
+ break;
+ }
+
+ if (rc)
+ goto out_airq;
+
+ /*
+ * Enable floating IRQs (with suppression after one IRQ). When using
+ * directed IRQs this enables the fallback path.
+ */
+ zpci_set_irq_ctrl(SIC_IRQ_MODE_SINGLE, PCI_ISC);
+
+ return 0;
out_airq:
unregister_adapter_interrupt(&zpci_airq);
out:
@@ -230,7 +467,15 @@ out:
void __init zpci_irq_exit(void)
{
- airq_iv_release(zpci_sbv);
+ unsigned int cpu;
+
+ if (irq_delivery == DIRECTED) {
+ for_each_possible_cpu(cpu) {
+ airq_iv_release(zpci_ibv[cpu]);
+ }
+ }
kfree(zpci_ibv);
+ if (zpci_sbv)
+ airq_iv_release(zpci_sbv);
unregister_adapter_interrupt(&zpci_airq);
}