From: Vasant Hegde <vasant.hegde@amd.com>
To: Jerry Snitselaar <jsnitsel@redhat.com>
Cc: iommu@lists.linux.dev, joro@8bytes.org,
suravee.suthikulpanit@amd.com, joao.m.martins@oracle.com
Subject: Re: [PATCH 1/2] iommu/amd: Add separate interrupt handler for PPR and GA log
Date: Wed, 14 Jun 2023 14:29:22 +0530 [thread overview]
Message-ID: <14e5691d-1914-031e-8a8c-5a18c62fc94d@amd.com> (raw)
In-Reply-To: <nm5ndwb7flwfa7tmgekubndrtmrd6brnjz77n2ub3iuoyyzcmv@msgzwfm3r4q5>
Hi Jerry,
On 6/14/2023 3:08 AM, Jerry Snitselaar wrote:
> On Fri, Jun 09, 2023 at 10:20:24AM +0000, Vasant Hegde wrote:
>> The AMD IOMMU has three different logs (Event, PPR and GA) and it can be
>> configured to send separate interrupt for each log type.
>> - Event log is used whenever IOMMU reports events like IO_PAGE_FAULT,
>> TLB_INV_TIMEOUT, etc,. During normal system operation this log is not
>> used actively.
>>
>> - GA log is used to record device interrupt requests that could not be
>> immediately delivered to the target virtual processor due the fact the
>> target was not running. This is actively used when we do device
>> passthrough to AVIC enabled guest.
>>
>> - PPR log is used to service the page fault request from device in Shared
>> Virtual Addressing (SVA) mode where page table is shared by CPU and
>> device. In this mode it will generate PPR interrupt frequently.
>>
>> Currently we have single interrupt to handle all three logs. GA log and
>> PPR log usage is increasing. Hence, split interrupt handler thread
>> into three separate interrupt handler function. Following patch enables
>> separate interrupt for PPR and GA Log.
>>
>> Signed-off-by: Vasant Hegde <vasant.hegde@amd.com>
>> ---
>> drivers/iommu/amd/amd_iommu.h | 3 ++
>> drivers/iommu/amd/iommu.c | 98 +++++++++++++++++++----------------
>> 2 files changed, 57 insertions(+), 44 deletions(-)
>>
>> diff --git a/drivers/iommu/amd/amd_iommu.h b/drivers/iommu/amd/amd_iommu.h
>> index 156f57b4f78c..e2857109e966 100644
>> --- a/drivers/iommu/amd/amd_iommu.h
>> +++ b/drivers/iommu/amd/amd_iommu.h
>> @@ -12,6 +12,9 @@
>> #include "amd_iommu_types.h"
>>
>> irqreturn_t amd_iommu_int_thread(int irq, void *data);
>> +irqreturn_t amd_iommu_int_thread_evtlog(int irq, void *data);
>> +irqreturn_t amd_iommu_int_thread_pprlog(int irq, void *data);
>> +irqreturn_t amd_iommu_int_thread_galog(int irq, void *data);
>> irqreturn_t amd_iommu_int_handler(int irq, void *data);
>> void amd_iommu_apply_erratum_63(struct amd_iommu *iommu, u16 devid);
>> void amd_iommu_restart_event_logging(struct amd_iommu *iommu);
>> diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
>> index 3c179d548ecd..d427f7e3b869 100644
>> --- a/drivers/iommu/amd/iommu.c
>> +++ b/drivers/iommu/amd/iommu.c
>> @@ -841,57 +841,23 @@ static inline void
>> amd_iommu_set_pci_msi_domain(struct device *dev, struct amd_iommu *iommu) { }
>> #endif /* !CONFIG_IRQ_REMAP */
>>
>> -#define AMD_IOMMU_INT_MASK \
>> - (MMIO_STATUS_EVT_OVERFLOW_MASK | \
>> - MMIO_STATUS_EVT_INT_MASK | \
>> - MMIO_STATUS_PPR_OVERFLOW_MASK | \
>> - MMIO_STATUS_PPR_INT_MASK | \
>> - MMIO_STATUS_GALOG_OVERFLOW_MASK | \
>> - MMIO_STATUS_GALOG_INT_MASK)
>> -
>> -irqreturn_t amd_iommu_int_thread(int irq, void *data)
>> +static void amd_iommu_handle_irq(void *data, u32 int_mask, u32 overflow_mask,
>> + void (*int_handler)(struct amd_iommu *),
>> + void (*overflow_handler)(struct amd_iommu *))
>> {
>> struct amd_iommu *iommu = (struct amd_iommu *) data;
>> u32 status = readl(iommu->mmio_base + MMIO_STATUS_OFFSET);
>> + u32 mask = int_mask | overflow_mask;
>>
>> - while (status & AMD_IOMMU_INT_MASK) {
>> + while (status & mask) {
>> /* Enable interrupt sources again */
>> - writel(AMD_IOMMU_INT_MASK,
>> - iommu->mmio_base + MMIO_STATUS_OFFSET);
>> -
>> - if (status & MMIO_STATUS_EVT_INT_MASK) {
>> - pr_devel("Processing IOMMU Event Log\n");
>> - iommu_poll_events(iommu);
>> - }
>> -
>> - if (status & (MMIO_STATUS_PPR_INT_MASK |
>> - MMIO_STATUS_PPR_OVERFLOW_MASK)) {
>> - pr_devel("Processing IOMMU PPR Log\n");
>> - iommu_poll_ppr_log(iommu);
>> - }
>> + writel(mask, iommu->mmio_base + MMIO_STATUS_OFFSET);
>>
>> - if (status & MMIO_STATUS_PPR_OVERFLOW_MASK) {
>> - pr_info_ratelimited("IOMMU PPR log overflow\n");
>> - amd_iommu_restart_ppr_log(iommu);
>> - }
>> + if (int_handler)
>> + int_handler(iommu);
>>
>> -#ifdef CONFIG_IRQ_REMAP
>> - if (status & (MMIO_STATUS_GALOG_INT_MASK |
>> - MMIO_STATUS_GALOG_OVERFLOW_MASK)) {
>> - pr_devel("Processing IOMMU GA Log\n");
>> - iommu_poll_ga_log(iommu);
>> - }
>> -
>> - if (status & MMIO_STATUS_GALOG_OVERFLOW_MASK) {
>> - pr_info_ratelimited("IOMMU GA Log overflow\n");
>> - amd_iommu_restart_ga_log(iommu);
>> - }
>> -#endif
>> -
>> - if (status & MMIO_STATUS_EVT_OVERFLOW_MASK) {
>> - pr_info_ratelimited("IOMMU event log overflow\n");
>> - amd_iommu_restart_event_logging(iommu);
>> - }
>> + if ((status & overflow_mask) && overflow_handler)
>> + overflow_handler(iommu);
>>
>> /*
>> * Hardware bug: ERBT1312
>> @@ -908,6 +874,50 @@ irqreturn_t amd_iommu_int_thread(int irq, void *data)
>> */
>> status = readl(iommu->mmio_base + MMIO_STATUS_OFFSET);
>> }
>> +}
>> +
>> +irqreturn_t amd_iommu_int_thread_evtlog(int irq, void *data)
>> +{
>> + pr_devel("Processing IOMMU Event Log\n");
>> + amd_iommu_handle_irq(data, MMIO_STATUS_EVT_INT_MASK,
>> + MMIO_STATUS_EVT_OVERFLOW_MASK,
>> + iommu_poll_events, amd_iommu_restart_event_logging);
>> +
>> + return IRQ_HANDLED;
>> +}
>> +
>> +irqreturn_t amd_iommu_int_thread_pprlog(int irq, void *data)
>> +{
>> + pr_devel("Processing IOMMU PPR Log\n");
>> + amd_iommu_handle_irq(data, MMIO_STATUS_PPR_INT_MASK,
>> + MMIO_STATUS_PPR_OVERFLOW_MASK,
>> + iommu_poll_ppr_log, amd_iommu_restart_ppr_log);
>> +
>> + return IRQ_HANDLED;
>> +}
>> +
>> +irqreturn_t amd_iommu_int_thread_galog(int irq, void *data)
>> +{
>> +
>> + pr_devel("Processing IOMMU GA Log\n");
>> +#ifdef CONFIG_IRQ_REMAP
>> + amd_iommu_handle_irq(data, MMIO_STATUS_GALOG_INT_MASK,
>> + MMIO_STATUS_GALOG_OVERFLOW_MASK,
>> + iommu_poll_ga_log, amd_iommu_restart_ga_log);
>> +#else
>> + amd_iommu_handle_irq(data, MMIO_STATUS_GALOG_INT_MASK,
>> + MMIO_STATUS_GALOG_OVERFLOW_MASK, NULL, NULL);
>> +#endif
>
> Is the else needed here, since GAEn will only get set if CONFIG_IRQ_REMAP is enabled?
In the current flow we clear the the interrupt bits (including GALOG_INT and
GALOG_OVERFLOW) first and then if IRQ_REMAP is enabled then we process GA log.
To match that flow I have added else part.
But you are right. We enable GALOG only when IRQ_REMAP is enabled.
Probably I can drop else part.
-Vasant
next prev parent reply other threads:[~2023-06-14 8:59 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-06-09 10:20 [PATCH 0/2] iommu/amd: Interrupt handling improvements Vasant Hegde
2023-06-09 10:20 ` [PATCH 1/2] iommu/amd: Add separate interrupt handler for PPR and GA log Vasant Hegde
2023-06-13 21:38 ` Jerry Snitselaar
2023-06-14 8:59 ` Vasant Hegde [this message]
2023-06-20 15:01 ` Joao Martins
2023-06-20 16:16 ` Vasant Hegde
2023-06-09 10:20 ` [PATCH 2/2] iommu/amd: Enable separate interrupt " Vasant Hegde
2023-06-14 17:18 ` Jerry Snitselaar
2023-06-19 10:16 ` Vasant Hegde
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=14e5691d-1914-031e-8a8c-5a18c62fc94d@amd.com \
--to=vasant.hegde@amd.com \
--cc=iommu@lists.linux.dev \
--cc=joao.m.martins@oracle.com \
--cc=joro@8bytes.org \
--cc=jsnitsel@redhat.com \
--cc=suravee.suthikulpanit@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).