genirq/msi: Make interrupt allocation less convoluted

There is no real reason to do several loops over the MSI descriptors
instead of just doing one loop. In case of an error everything is undone
anyway so it does not matter whether it's a partial or a full rollback.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Michael Kelley <mikelley@microsoft.com>
Tested-by: Nishanth Menon <nm@ti.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20211206210749.010234767@linutronix.de
This commit is contained in:
Thomas Gleixner 2021-12-06 23:51:44 +01:00
parent a80713fea3
commit ef8dd01538
3 changed files with 69 additions and 67 deletions

View File

@ -216,7 +216,6 @@ ForEachMacros:
- 'for_each_migratetype_order' - 'for_each_migratetype_order'
- 'for_each_msi_entry' - 'for_each_msi_entry'
- 'for_each_msi_entry_safe' - 'for_each_msi_entry_safe'
- 'for_each_msi_vector'
- 'for_each_net' - 'for_each_net'
- 'for_each_net_continue_reverse' - 'for_each_net_continue_reverse'
- 'for_each_netdev' - 'for_each_netdev'

View File

@ -206,12 +206,6 @@ struct msi_desc *msi_next_desc(struct device *dev, enum msi_desc_filter filter);
list_for_each_entry((desc), dev_to_msi_list((dev)), list) list_for_each_entry((desc), dev_to_msi_list((dev)), list)
#define for_each_msi_entry_safe(desc, tmp, dev) \ #define for_each_msi_entry_safe(desc, tmp, dev) \
list_for_each_entry_safe((desc), (tmp), dev_to_msi_list((dev)), list) list_for_each_entry_safe((desc), (tmp), dev_to_msi_list((dev)), list)
#define for_each_msi_vector(desc, __irq, dev) \
for_each_msi_entry((desc), (dev)) \
if ((desc)->irq) \
for (__irq = (desc)->irq; \
__irq < ((desc)->irq + (desc)->nvec_used); \
__irq++)
#ifdef CONFIG_IRQ_MSI_IOMMU #ifdef CONFIG_IRQ_MSI_IOMMU
static inline const void *msi_desc_get_iommu_cookie(struct msi_desc *desc) static inline const void *msi_desc_get_iommu_cookie(struct msi_desc *desc)

View File

@ -828,23 +828,74 @@ static int msi_handle_pci_fail(struct irq_domain *domain, struct msi_desc *desc,
return allocated ? allocated : -ENOSPC; return allocated ? allocated : -ENOSPC;
} }
#define VIRQ_CAN_RESERVE 0x01
#define VIRQ_ACTIVATE 0x02
#define VIRQ_NOMASK_QUIRK 0x04
static int msi_init_virq(struct irq_domain *domain, int virq, unsigned int vflags)
{
struct irq_data *irqd = irq_domain_get_irq_data(domain, virq);
int ret;
if (!(vflags & VIRQ_CAN_RESERVE)) {
irqd_clr_can_reserve(irqd);
if (vflags & VIRQ_NOMASK_QUIRK)
irqd_set_msi_nomask_quirk(irqd);
}
if (!(vflags & VIRQ_ACTIVATE))
return 0;
ret = irq_domain_activate_irq(irqd, vflags & VIRQ_CAN_RESERVE);
if (ret)
return ret;
/*
* If the interrupt uses reservation mode, clear the activated bit
* so request_irq() will assign the final vector.
*/
if (vflags & VIRQ_CAN_RESERVE)
irqd_clr_activated(irqd);
return 0;
}
int __msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, int __msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev,
int nvec) int nvec)
{ {
struct msi_domain_info *info = domain->host_data; struct msi_domain_info *info = domain->host_data;
struct msi_domain_ops *ops = info->ops; struct msi_domain_ops *ops = info->ops;
struct irq_data *irq_data;
struct msi_desc *desc;
msi_alloc_info_t arg = { }; msi_alloc_info_t arg = { };
unsigned int vflags = 0;
struct msi_desc *desc;
int allocated = 0; int allocated = 0;
int i, ret, virq; int i, ret, virq;
bool can_reserve;
ret = msi_domain_prepare_irqs(domain, dev, nvec, &arg); ret = msi_domain_prepare_irqs(domain, dev, nvec, &arg);
if (ret) if (ret)
return ret; return ret;
for_each_msi_entry(desc, dev) { /*
* This flag is set by the PCI layer as we need to activate
* the MSI entries before the PCI layer enables MSI in the
* card. Otherwise the card latches a random msi message.
*/
if (info->flags & MSI_FLAG_ACTIVATE_EARLY)
vflags |= VIRQ_ACTIVATE;
/*
* Interrupt can use a reserved vector and will not occupy
* a real device vector until the interrupt is requested.
*/
if (msi_check_reservation_mode(domain, info, dev)) {
vflags |= VIRQ_CAN_RESERVE;
/*
* MSI affinity setting requires a special quirk (X86) when
* reservation mode is active.
*/
if (domain->flags & IRQ_DOMAIN_MSI_NOMASK_QUIRK)
vflags |= VIRQ_NOMASK_QUIRK;
}
msi_for_each_desc(desc, dev, MSI_DESC_NOTASSOCIATED) {
ops->set_desc(&arg, desc); ops->set_desc(&arg, desc);
virq = __irq_domain_alloc_irqs(domain, -1, desc->nvec_used, virq = __irq_domain_alloc_irqs(domain, -1, desc->nvec_used,
@ -856,49 +907,12 @@ int __msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev,
for (i = 0; i < desc->nvec_used; i++) { for (i = 0; i < desc->nvec_used; i++) {
irq_set_msi_desc_off(virq, i, desc); irq_set_msi_desc_off(virq, i, desc);
irq_debugfs_copy_devname(virq + i, dev); irq_debugfs_copy_devname(virq + i, dev);
ret = msi_init_virq(domain, virq + i, vflags);
if (ret)
return ret;
} }
allocated++; allocated++;
} }
can_reserve = msi_check_reservation_mode(domain, info, dev);
/*
* This flag is set by the PCI layer as we need to activate
* the MSI entries before the PCI layer enables MSI in the
* card. Otherwise the card latches a random msi message.
*/
if (!(info->flags & MSI_FLAG_ACTIVATE_EARLY))
goto skip_activate;
for_each_msi_vector(desc, i, dev) {
if (desc->irq == i) {
virq = desc->irq;
dev_dbg(dev, "irq [%d-%d] for MSI\n",
virq, virq + desc->nvec_used - 1);
}
irq_data = irq_domain_get_irq_data(domain, i);
if (!can_reserve) {
irqd_clr_can_reserve(irq_data);
if (domain->flags & IRQ_DOMAIN_MSI_NOMASK_QUIRK)
irqd_set_msi_nomask_quirk(irq_data);
}
ret = irq_domain_activate_irq(irq_data, can_reserve);
if (ret)
return ret;
}
skip_activate:
/*
* If these interrupts use reservation mode, clear the activated bit
* so request_irq() will assign the final vector.
*/
if (can_reserve) {
for_each_msi_vector(desc, i, dev) {
irq_data = irq_domain_get_irq_data(domain, i);
irqd_clr_activated(irq_data);
}
}
return 0; return 0;
} }
@ -976,26 +990,21 @@ int msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, int nve
void __msi_domain_free_irqs(struct irq_domain *domain, struct device *dev) void __msi_domain_free_irqs(struct irq_domain *domain, struct device *dev)
{ {
struct irq_data *irq_data; struct irq_data *irqd;
struct msi_desc *desc; struct msi_desc *desc;
int i; int i;
for_each_msi_vector(desc, i, dev) { /* Only handle MSI entries which have an interrupt associated */
irq_data = irq_domain_get_irq_data(domain, i); msi_for_each_desc(desc, dev, MSI_DESC_ASSOCIATED) {
if (irqd_is_activated(irq_data)) /* Make sure all interrupts are deactivated */
irq_domain_deactivate_irq(irq_data); for (i = 0; i < desc->nvec_used; i++) {
} irqd = irq_domain_get_irq_data(domain, desc->irq + i);
if (irqd && irqd_is_activated(irqd))
for_each_msi_entry(desc, dev) { irq_domain_deactivate_irq(irqd);
/*
* We might have failed to allocate an MSI early
* enough that there is no IRQ associated to this
* entry. If that's the case, don't do anything.
*/
if (desc->irq) {
irq_domain_free_irqs(desc->irq, desc->nvec_used);
desc->irq = 0;
} }
irq_domain_free_irqs(desc->irq, desc->nvec_used);
desc->irq = 0;
} }
} }