Re: [PATCH v4 2/5] cxl/trace: Add TRACE support for CXL media-error records

From: Alison Schofield
Date: Thu Jan 12 2023 - 11:52:55 EST


On Thu, Jan 12, 2023 at 11:16:52AM +0000, Jonathan Cameron wrote:
> On Thu, 15 Dec 2022 13:17:44 -0800
> alison.schofield@xxxxxxxxx wrote:
>
> > From: Alison Schofield <alison.schofield@xxxxxxxxx>
> >
> > CXL devices may support the retrieval of a device poison list.
> > Add a new trace event that the CXL subsystem may use to log
> > the media-error records returned in the poison list.
> >
> > Log each media-error record as a trace event of type 'cxl_poison'.
> >
> > When the poison list is requested by region, include the region name
> > and uuid in the trace event.
> >
> > Reviewed-by: Jonathan Cameron <Jonathan.Cameron@xxxxxxxxxx>
> > Signed-off-by: Alison Schofield <alison.schofield@xxxxxxxxx>
>
> Hi Alison,
>
> I'm wondering a bit if it makes sense to log/trace the MORE flag for each
> record. That has some unusual semantics. Like the other flags it's a
> characteristic of the underlying Get Poison List Output Payload, not a
> particularly Poison list entry. Unlike overflow and scanning which are
> reasonably a characteristic of each poison record in a given set read
> in one command, MORE is not.
> Imagine the following.
>
> Read first N records, MORE set so we have that flag in all the trace
> records.
> Read next M records, MORE not set as list is less than N+M records.
>
> Now userspace tooling has no idea about the mapping to underlying
> mailbox commands. So it sees this bit set for first N which is fine
> as there are more records following, but for the next M - 1 it might
> reasonably expect to see MORE set (as more records are getting traced)
> but it doesn't see it for any of them.
>
> That seems less than intuitive.
>
> I'm not sure what to suggest... Seems wrong to 'make up a MORE flag' for
> those M-1 records. Does it make sense to expose this flag at all? In
> what way is it useful?
>
> Jonathan

Thanks Jonathan -

The FLAGs field today simply reflects the contents of the flags
field of the payload, which can be MORE, OVERFLOW, SCANNING.

Understand that 'MORE' seems useless to userspace however, I lean
towards giving userspace the complete truth, and letting userspace
ignore what is of no interest.

Perhaps a user wants to explore the boundaries of a devices
poison handling capacity. ie...use inject and get_poison and
see when the MORE flag appears. I don't know if it may appear
before the inject_max is reached.

So - I say keep MORE as is.

This discussion triggers another thought...

ATM, a devices max_mer - Max Media Error Records is not exposed in
sysfs, and this conversation makes me think it should be. It also,
comes to mind because I am exposing the inject_poison_limit to sysfs.

Users wanting to explore poison capabilities would be interested in
max_mer.

I'll add the max_mer to syfs in next rev.

Alison

>
>
> > ---
> > drivers/cxl/core/mbox.c | 6 ++-
> > drivers/cxl/core/trace.h | 83 ++++++++++++++++++++++++++++++++++++++++
> > 2 files changed, 88 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/cxl/core/mbox.c b/drivers/cxl/core/mbox.c
> > index dfe24a2adfdb..c345af8a4afd 100644
> > --- a/drivers/cxl/core/mbox.c
> > +++ b/drivers/cxl/core/mbox.c
> > @@ -10,6 +10,7 @@
> > #include <cxl.h>
> >
> > #include "core.h"
> > +#include "trace.h"
> >
> > static bool cxl_raw_allow_all;
> >
> > @@ -899,7 +900,10 @@ int cxl_mem_get_poison(struct cxl_memdev *cxlmd, u64 offset, u64 len,
> > if (rc)
> > break;
> >
> > - /* TODO TRACE the media error records */
> > + for (int i = 0; i < le16_to_cpu(po->count); i++)
> > + trace_cxl_poison(cxlmd, to_pci_dev(cxlds->dev),
> > + cxlr, &po->record[i], po->flags,
> > + po->overflow_t);
> >
> > /* Protect against an uncleared _FLAG_MORE */
> > nr_records = nr_records + le16_to_cpu(po->count);
> > diff --git a/drivers/cxl/core/trace.h b/drivers/cxl/core/trace.h
> > index 20ca2fe2ca8e..c7958311ce5f 100644
> > --- a/drivers/cxl/core/trace.h
> > +++ b/drivers/cxl/core/trace.h
> > @@ -8,6 +8,9 @@
> >
> > #include <cxl.h>
> > #include <linux/tracepoint.h>
> > +#include <linux/pci.h>
> > +
> > +#include <cxlmem.h>
> >
> > #define CXL_RAS_UC_CACHE_DATA_PARITY BIT(0)
> > #define CXL_RAS_UC_CACHE_ADDR_PARITY BIT(1)
> > @@ -103,6 +106,86 @@ TRACE_EVENT(cxl_aer_correctable_error,
> > )
> > );
> >
> > +#define __show_poison_source(source) \
> > + __print_symbolic(source, \
> > + { CXL_POISON_SOURCE_UNKNOWN, "Unknown" }, \
> > + { CXL_POISON_SOURCE_EXTERNAL, "External" }, \
> > + { CXL_POISON_SOURCE_INTERNAL, "Internal" }, \
> > + { CXL_POISON_SOURCE_INJECTED, "Injected" }, \
> > + { CXL_POISON_SOURCE_VENDOR, "Vendor" })
> > +
> > +#define show_poison_source(source) \
> > + (((source > CXL_POISON_SOURCE_INJECTED) && \
> > + (source != CXL_POISON_SOURCE_VENDOR)) ? "Reserved" \
> > + : __show_poison_source(source))
> > +
> > +#define show_poison_flags(flags) \
> > + __print_flags(flags, "|", \
> > + { CXL_POISON_FLAG_MORE, "More" }, \
> > + { CXL_POISON_FLAG_OVERFLOW, "Overflow" }, \
> > + { CXL_POISON_FLAG_SCANNING, "Scanning" })
> > +
> > +#define __cxl_poison_addr(record) \
> > + (le64_to_cpu(record->address))
> > +#define cxl_poison_record_dpa(record) \
> > + (__cxl_poison_addr(record) & CXL_POISON_START_MASK)
> > +#define cxl_poison_record_source(record) \
> > + (__cxl_poison_addr(record) & CXL_POISON_SOURCE_MASK)
> > +#define cxl_poison_record_length(record) \
> > + (le32_to_cpu(record->length) * CXL_POISON_LEN_MULT)
> > +#define cxl_poison_overflow(flags, time) \
> > + (flags & CXL_POISON_FLAG_OVERFLOW ? le64_to_cpu(time) : 0)
> > +
> > +TRACE_EVENT(cxl_poison,
> > +
> > + TP_PROTO(struct cxl_memdev *memdev, const struct pci_dev *pcidev,
> > + struct cxl_region *region,
> > + const struct cxl_poison_record *record,
> > + u8 flags, __le64 overflow_t),
> > +
> > + TP_ARGS(memdev, pcidev, region, record, flags, overflow_t),
> > +
> > + TP_STRUCT__entry(
> > + __string(memdev, dev_name(&memdev->dev))
> > + __string(pcidev, dev_name(&pcidev->dev))
> > + __string(region, region)
> > + __field(u64, overflow_t)
> > + __field(u64, dpa)
> > + __field(u32, length)
> > + __array(char, uuid, 16)
> > + __field(u8, source)
> > + __field(u8, flags)
> > + ),
> > +
> > + TP_fast_assign(
> > + __assign_str(memdev, dev_name(&memdev->dev));
> > + __assign_str(pcidev, dev_name(&pcidev->dev));
> > + __entry->overflow_t = cxl_poison_overflow(flags, overflow_t);
> > + __entry->dpa = cxl_poison_record_dpa(record);
> > + __entry->length = cxl_poison_record_length(record);
> > + __entry->source = cxl_poison_record_source(record);
> > + __entry->flags = flags;
> > + if (region) {
> > + __assign_str(region, dev_name(&region->dev));
> > + memcpy(__entry->uuid, &region->params.uuid, 16);
> > + } else {
> > + __assign_str(region, "");
> > + memset(__entry->uuid, 0, 16);
> > + }
> > + ),
> > +
> > + TP_printk("memdev=%s pcidev=%s region=%s region_uuid=%pU dpa=0x%llx length=0x%x source=%s flags=%s overflow_time=%llu",
> > + __get_str(memdev),
> > + __get_str(pcidev),
> > + __get_str(region),
> > + __entry->uuid,
> > + __entry->dpa,
> > + __entry->length,
> > + show_poison_source(__entry->source),
> > + show_poison_flags(__entry->flags),
> > + __entry->overflow_t)
> > +);
> > +
> > #endif /* _CXL_EVENTS_H */
> >
> > #define TRACE_INCLUDE_FILE trace
>