XOR Math Fixups: translation & position
[email protected]
maintainerv2 cover XOR Math Fixups: translation & position
From: Alison Schofield <[email protected]>
Changes in v2:
- Rebase on cxl/next
- Remove 'misconfigured root decoder' warn in cxl_region_attach_position() (Dan)
- Comment the implementation of CXL Spec's XORALLBITS in cxl_xor_restore (Dan)
- Prepend new Patch 1 to rename cxl_trace_hpa()-> cxl_translate() (Dan)
- Simplify the setting of position bit in cxl_xor_restore() (Dan)
- Collapse the helper restore_xor_pos() into cxl_xor_restore()
- Use host bridge ways, not region ways, in cxl_xor_restore().
- Update 'override' comment in cxl_translate() (Dan)
- Clarify via renames (Dan)
cxl_addr_trans_fn -> cxl_translate_fn cxl_xor_trans() -> cxl_xor_translate()
Link to v1:
https://lore.kernel.org/[email protected]/
Begin cover letter:
Rather than repeat the individual patch commit message content, let me describe the flow of this set:
Patch 1: Rename an existing fn - cxl_trace_hpa()-> cxl_translate()
Patch 2: cxl/acpi: Restore XOR'd position bits during address translation The problem fixed in this patch, bad HPA translations with XOR math, came to my attention recently. Patch 1 can stand alone, but since that discovery also shed light on how to repair an issue with calculating positions in interleave sets (Patch 2,3) they are presented together.
Patch 3 & Patch 4 are paired. Patch 3 presents the new method for verifying a target position in the list and Patch 3 removes the old method. These could be squashed.
FYI - the reason I don't present the code removal first is because I think it is easier to read the diff if I leave in the old root decoder call back setup for calc_hb, insert the new call back along the same path, and then rip out the defunct calc_hb. That's the way I created the patchset and it may be an easier way for reviewers to follow along with the root decoder callback setup.
Alison Schofield (4):
cxl/core: Rename cxl_trace_hpa() to cxl_translate() cxl/acpi: Restore XOR'd position bits during address translation cxl/region: Verify target positions using the ordered target list
cxl: Remove defunct code calculating host bridge target positions
drivers/cxl/acpi.c | 80 ++++++++++++++++----------------------- drivers/cxl/core/core.h | 4 +- drivers/cxl/core/mbox.c | 2 +- drivers/cxl/core/port.c | 21 ++-------- drivers/cxl/core/region.c | 12 +++++- drivers/cxl/core/trace.h | 2 +- drivers/cxl/cxl.h | 10 ++--- 7 files changed, 54 insertions(+), 77 deletions(-)
base-commit: d99f13843237cf9dbdc1bd873a901662b4aee16f
Signature · 1 line
2.37.3
[email protected]
maintainerPATCH v2 1/4 cxl/core: Rename cxl_trace_hpa() to cxl_translate()
From: Alison Schofield <[email protected]>
Although cxl_trace_hpa() is used to populate TRACE EVENTs with HPA addresses, the work it performs is a translation (dpa->hpa), not a trace. Rename it. Since this is the only translate work in CXL, drop the _hpa suffix in the rename.
--- drivers/cxl/core/core.h | 4 ++-- drivers/cxl/core/mbox.c | 2 +- drivers/cxl/core/region.c | 2 +- drivers/cxl/core/trace.h | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-)
drivers/cxl/core/core.h +2 -2
@@ -28,12 +28,12 @@ int cxl_region_init(void);void cxl_region_exit (void );int cxl_get_poison_by_endpoint (struct cxl_port *port);struct cxl_region *cxl_dpa_to_region (const struct cxl_memdev *cxlmd,u64 dpa); -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa);#else static inline u64 -cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) +cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {return ULLONG_MAX ; }
drivers/cxl/core/mbox.c +1 -1
@@ -878,7 +878,7 @@ void cxl_event_trace_record(const struct cxl_memdev *cxlmd, dpa =le64_to_cpu (evt->common.phys_addr) &CXL_DPA_MASK ; cxlr =cxl_dpa_to_region (cxlmd, dpa);if (cxlr) - hpa =cxl_trace_hpa (cxlr, cxlmd, dpa); + hpa =cxl_translate (cxlr, cxlmd, dpa);if (event_type ==CXL_CPER_EVENT_GEN_MEDIA ) trace_cxl_general_media(cxlmd, type, cxlr, hpa,
drivers/cxl/core/region.c +1 -1
@@ -2797,7 +2797,7 @@ static u64 cxl_dpa_to_hpa(u64 dpa, struct cxl_region *cxlr,return hpa; } -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {struct cxl_region_params *p = &cxlr->params;
drivers/cxl/core/trace.h +1 -1
@@ -704,7 +704,7 @@ TRACE_EVENT(cxl_poison,if (cxlr) {__assign_str (region,dev_name (&cxlr->dev));memcpy (__entry->uuid, &cxlr->params.uuid,16 ); - __entry->hpa =cxl_trace_hpa (cxlr, cxlmd, + __entry->hpa =cxl_translate (cxlr, cxlmd, __entry->dpa); }else {__assign_str (region,"" );
Signature · 1 line
2.37.3
[email protected]
maintainerPATCH v2 2/4 cxl/acpi: Restore XOR'd position bits during address translation
In reply to [email protected]From: Alison Schofield <[email protected]>
When a CXL region is created in a CXL Window (CFMWS) that uses XOR interleave arithmetic XOR maps are applied during the HPA->DPA translation. The XOR function changes the interleave selector bit (aka position bit) in the HPA thereby varying which host bridge services an HPA. The purpose is to minimize hot spots thereby improving performance.
When a device reports a DPA in events such as poison, general_media, and dram, the driver translates that DPA back to an HPA. Presently, the CXL driver translation only considers the modulo position and will report the wrong HPA for XOR configured CFMWS's.
Add a helper function that restores the XOR'd bits during DPA->HPA address translation. Plumb a root decoder callback to the new helper when XOR interleave arithmetic is in use. For MODULO arithmetic, just let the callback be NULL - as in no extra work required.
--- drivers/cxl/acpi.c | 48 ++++++++++++++++++++++++++++++++++++--- drivers/cxl/core/port.c | 5 +++- drivers/cxl/core/region.c | 5 ++++ drivers/cxl/cxl.h | 6 ++++- 4 files changed, 59 insertions(+), 5 deletions(-)
drivers/cxl/acpi.c +45 -3
@@ -74,6 +74,43 @@ static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos)return cxlrd->cxlsd.target[n]; } +static u64 cxl_xor_translate (struct cxl_root_decoder *cxlrd,u64 hpa) +{ +struct cxl_cxims_data *cximsd = cxlrd->platform_data; +int hbiw = cxlrd->cxlsd.nr_targets; +u64 val; +int pos; + +/* No xormaps for host bridge interleave ways of 1 or 3 */ +if (hbiw ==1 || hbiw ==3 ) +return hpa; + +/* +* For root decoders using xormaps (hbiw: 2,4,6,8,12,16) restore +* the position bit to its value before the xormap was applied at +* HPA->DPA translation. +* +* pos is the lowest set bit in an XORMAP +* val is the XORALLBITS(HPA & XORMAP) +* +* XORALLBITS: The CXL spec (3.1 Table 9-22) defines XORALLBITS +* as an operation that outputs a single bit by XORing all the +* bits in the input (hpa & xormap). Implement XORALLBITS using +* hweight64(). If the hamming weight is even the XOR of those +* bits results in 0, if odd the XOR result is 1. +*/ + +for (int i =0 ; i < cximsd->nr_maps; i++) { +if (!cximsd->xormaps[i]) +continue ; + pos =__ffs (cximsd->xormaps[i]); + val = (hweight64 (hpa & cximsd->xormaps[i]) &1 ); + hpa = (hpa & ~(1ULL << pos)) | (val << pos); + } + +return hpa; +} +struct cxl_cxims_context {struct device *dev;struct cxl_root_decoder *cxlrd;
@@ -362,6 +399,7 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,struct cxl_cxims_context cxims_ctx;struct device *dev = ctx->dev;cxl_calc_hb_fn cxl_calc_hb; +cxl_translate_fn translate;struct cxl_decoder *cxld;unsigned int ways, i, ig;int rc;
@@ -389,13 +427,17 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,if (rc)return rc; -if (cfmws->interleave_arithmetic ==ACPI_CEDT_CFMWS_ARITHMETIC_MODULO ) +if (cfmws->interleave_arithmetic ==ACPI_CEDT_CFMWS_ARITHMETIC_MODULO ) { cxl_calc_hb = cxl_hb_modulo; -else + translate =NULL ; + + }else { cxl_calc_hb = cxl_hb_xor; + translate = cxl_xor_translate; + } struct cxl_root_decoder *cxlrd __free(put_cxlrd) = -cxl_root_decoder_alloc (root_port, ways, cxl_calc_hb); +cxl_root_decoder_alloc (root_port, ways, cxl_calc_hb, translate);if (IS_ERR (cxlrd))return PTR_ERR (cxlrd);
drivers/cxl/core/port.c +4 -1
@@ -1808,6 +1808,7 @@ static int cxl_switch_decoder_init(struct cxl_port *port,* @port: owning CXL root of this decoder * @nr_targets: static number of downstream targets * @calc_hb: which host bridge covers the n'th position by granularity +* @translate: decoder specific address translation function * * Return: A new cxl decoder to be registered by cxl_decoder_add(). A * 'CXL root' decoder is one that decodes from a top-level / static platform
@@ -1816,7 +1817,8 @@ static int cxl_switch_decoder_init(struct cxl_port *port,*/ struct cxl_root_decoder *cxl_root_decoder_alloc (struct cxl_port *port,unsigned int nr_targets, -cxl_calc_hb_fn calc_hb) +cxl_calc_hb_fn calc_hb, +cxl_translate_fn translate) {struct cxl_root_decoder *cxlrd;struct cxl_switch_decoder *cxlsd;
@@ -1839,6 +1841,7 @@ struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port, } cxlrd->calc_hb = calc_hb; + cxlrd->translate = translate;mutex_init (&cxlrd->range_lock); cxld = &cxlsd->cxld;
drivers/cxl/core/region.c +5 -0
@@ -2752,6 +2752,7 @@ static bool cxl_is_hpa_in_range(u64 hpa, struct cxl_region *cxlr, int pos)static u64 cxl_dpa_to_hpa (u64 dpa,struct cxl_region *cxlr,struct cxl_endpoint_decoder *cxled) { +struct cxl_root_decoder *cxlrd =to_cxl_root_decoder (cxlr->dev.parent);u64 dpa_offset, hpa_offset, bits_upper, mask_upper, hpa;struct cxl_region_params *p = &cxlr->params;int pos = cxled->pos;
@@ -2791,6 +2792,10 @@ static u64 cxl_dpa_to_hpa(u64 dpa, struct cxl_region *cxlr,/* Apply the hpa_offset to the region base address */ hpa = hpa_offset + p->res->start; +/* Root decoder translation overrides typical modulo decode */ +if (cxlrd->translate) + hpa = cxlrd->translate (cxlrd, hpa); +if (!cxl_is_hpa_in_range (hpa, cxlr, cxled->pos))return ULLONG_MAX ;
drivers/cxl/cxl.h +5 -1
@@ -434,12 +434,14 @@ struct cxl_switch_decoder {struct cxl_root_decoder ;typedef struct cxl_dport *(*cxl_calc_hb_fn )(struct cxl_root_decoder *cxlrd,int pos); +typedef u64 (*cxl_translate_fn )(struct cxl_root_decoder *cxlrd,u64 hpa);/** * struct cxl_root_decoder - Static platform CXL address decoder * @res: host / parent resource for region allocations * @region_id: region id for next region provisioning event * @calc_hb: which host bridge covers the n'th position by granularity +* @translate: decoder specific address translation function * @platform_data: platform specific configuration data * @range_lock: sync region autodiscovery by address range * @qos_class: QoS performance class cookie
@@ -449,6 +451,7 @@ struct cxl_root_decoder {struct resource *res;atomic_t region_id;cxl_calc_hb_fn calc_hb; +cxl_translate_fn translate;void *platform_data;struct mutex range_lock;int qos_class;
@@ -773,7 +776,8 @@ bool is_switch_decoder(struct device *dev);bool is_endpoint_decoder (struct device *dev);struct cxl_root_decoder *cxl_root_decoder_alloc (struct cxl_port *port,unsigned int nr_targets, -cxl_calc_hb_fn calc_hb); +cxl_calc_hb_fn calc_hb, +cxl_translate_fn translate);struct cxl_dport *cxl_hb_modulo (struct cxl_root_decoder *cxlrd,int pos);struct cxl_switch_decoder *cxl_switch_decoder_alloc (struct cxl_port *port,unsigned int nr_targets);
Signature · 1 line
2.37.3
[email protected]
maintainerPATCH v2 3/4 cxl/region: Verify target positions using the ordered target list
In reply to [email protected]From: Alison Schofield <[email protected]>
When a root decoder is configured the interleave target list is read from the BIOS populated CFMWS structure. Per the CXL spec 3.1 Table 9-22 the target list is in interleave order. The CXL driver populates its decoder target list in the same order and stores it in 'struct cxl_switch_decoder' field "@target: active ordered target list in current decoder configuration"
Given the promise of an ordered list, the driver can stop duplicating the work of BIOS and simply check target positions against the ordered list during region configuration.
The simplified check against the ordered list is presented here. A follow-on patch will remove the unused code.
For Modulo arithmetic this is not a fix, only a simplification. For XOR arithmetic this is a fix for HB IW of 3,6,12.
--- drivers/cxl/core/region.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/cxl/core/region.c +4 -1
@@ -1559,10 +1559,13 @@ static int cxl_region_attach_position(struct cxl_region *cxlr, const struct cxl_dport *dport, int pos) {struct cxl_memdev *cxlmd =cxled_to_memdev (cxled); +struct cxl_switch_decoder *cxlsd = &cxlrd->cxlsd; +struct cxl_decoder *cxld = &cxlsd->cxld; +int iw = cxld->interleave_ways;struct cxl_port *iter;int rc; -if (cxlrd->calc_hb (cxlrd, pos) != dport) { +if (dport != cxlrd->cxlsd.target[pos % iw]) {dev_dbg (&cxlr->dev,"%s:%s invalid target position for %s\n" ,dev_name (&cxlmd->dev),dev_name (&cxled->cxld.dev),dev_name (&cxlrd->cxlsd.cxld.dev));
Signature · 1 line
2.37.3
[email protected]
maintainerPATCH v2 4/4 cxl: Remove defunct code calculating host bridge target positions
In reply to [email protected]From: Alison Schofield <[email protected]>
The CXL Spec 3.1 Table 9-22 requires that the BIOS populate the CFMWS target list in interleave target order. This means that the calculations the CXL driver added to determine positions when XOR math is in use, along with the entire XOR vs Modulo call back setup is not needed.
A prior patch added a common method to verify positions.
Remove the now unused code related to the cxl_calc_hb_fn.
--- drivers/cxl/acpi.c | 62 ++--------------------------------------- drivers/cxl/core/port.c | 18 ------------ drivers/cxl/cxl.h | 6 ---- 3 files changed, 3 insertions(+), 83 deletions(-)
drivers/cxl/acpi.c +3 -59
@@ -22,57 +22,6 @@ static const guid_t acpi_cxl_qtg_id_guid =GUID_INIT (0xF365F9A6 ,0xA7DE ,0x4071 ,0xA6 ,0x6A ,0xB4 ,0x0C ,0x0B ,0x4F ,0x8E ,0x52 ); -/* -* Find a targets entry (n) in the host bridge interleave list. -* CXL Specification 3.0 Table 9-22 -*/ -static int cxl_xor_calc_n (u64 hpa,struct cxl_cxims_data *cximsd,int iw, -int ig) -{ -int i =0 , n =0 ; -u8 eiw; - -/* IW: 2,4,6,8,12,16 begin building 'n' using xormaps */ -if (iw !=3 ) { -for (i =0 ; i < cximsd->nr_maps; i++) - n |= (hweight64 (hpa & cximsd->xormaps[i]) &1 ) << i; - } -/* IW: 3,6,12 add a modulo calculation to 'n' */ -if (!is_power_of_2 (iw)) { -if (ways_to_eiw (iw, &eiw)) -return -1 ; - hpa &=GENMASK_ULL (51 , eiw + ig); - n |=do_div (hpa,3 ) << i; - } -return n; -} - -static struct cxl_dport *cxl_hb_xor (struct cxl_root_decoder *cxlrd,int pos) -{ -struct cxl_cxims_data *cximsd = cxlrd->platform_data; -struct cxl_switch_decoder *cxlsd = &cxlrd->cxlsd; -struct cxl_decoder *cxld = &cxlsd->cxld; -int ig = cxld->interleave_granularity; -int iw = cxld->interleave_ways; -int n =0 ; -u64 hpa; - -if (dev_WARN_ONCE (&cxld->dev, - cxld->interleave_ways != cxlsd->nr_targets, -"misconfigured root decoder\n" )) -return NULL ; - - hpa = cxlrd->res->start + pos * ig; - -/* Entry (n) is 0 for no interleave (iw == 1) */ -if (iw !=1 ) - n =cxl_xor_calc_n (hpa, cximsd, iw, ig); - -if (n <0 ) -return NULL ; - -return cxlrd->cxlsd.target[n]; -}static u64 cxl_xor_translate (struct cxl_root_decoder *cxlrd,u64 hpa) {
@@ -398,7 +347,6 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,struct cxl_port *root_port = ctx->root_port;struct cxl_cxims_context cxims_ctx;struct device *dev = ctx->dev; -cxl_calc_hb_fn cxl_calc_hb;cxl_translate_fn translate;struct cxl_decoder *cxld;unsigned int ways, i, ig;
@@ -427,17 +375,13 @@ static int __cxl_parse_cfmws(struct acpi_cedt_cfmws *cfmws,if (rc)return rc; -if (cfmws->interleave_arithmetic ==ACPI_CEDT_CFMWS_ARITHMETIC_MODULO ) { - cxl_calc_hb = cxl_hb_modulo; +if (cfmws->interleave_arithmetic ==ACPI_CEDT_CFMWS_ARITHMETIC_MODULO ) translate =NULL ; - - }else { - cxl_calc_hb = cxl_hb_xor; +else translate = cxl_xor_translate; - } struct cxl_root_decoder *cxlrd __free(put_cxlrd) = -cxl_root_decoder_alloc (root_port, ways, cxl_calc_hb, translate); +cxl_root_decoder_alloc (root_port, ways, translate);if (IS_ERR (cxlrd))return PTR_ERR (cxlrd);
drivers/cxl/core/port.c +0 -18
@@ -1733,21 +1733,6 @@ static int decoder_populate_targets(struct cxl_switch_decoder *cxlsd,return 0 ; } -struct cxl_dport *cxl_hb_modulo (struct cxl_root_decoder *cxlrd,int pos) -{ -struct cxl_switch_decoder *cxlsd = &cxlrd->cxlsd; -struct cxl_decoder *cxld = &cxlsd->cxld; -int iw; - - iw = cxld->interleave_ways; -if (dev_WARN_ONCE (&cxld->dev, iw != cxlsd->nr_targets, -"misconfigured root decoder\n" )) -return NULL ; - -return cxlrd->cxlsd.target[pos % iw]; -} -EXPORT_SYMBOL_NS_GPL (cxl_hb_modulo,CXL ); -static struct lock_class_key cxl_decoder_key;/**
@@ -1807,7 +1792,6 @@ static int cxl_switch_decoder_init(struct cxl_port *port,* cxl_root_decoder_alloc - Allocate a root level decoder * @port: owning CXL root of this decoder * @nr_targets: static number of downstream targets -* @calc_hb: which host bridge covers the n'th position by granularity * @translate: decoder specific address translation function * * Return: A new cxl decoder to be registered by cxl_decoder_add(). A
@@ -1817,7 +1801,6 @@ static int cxl_switch_decoder_init(struct cxl_port *port,*/ struct cxl_root_decoder *cxl_root_decoder_alloc (struct cxl_port *port,unsigned int nr_targets, -cxl_calc_hb_fn calc_hb,cxl_translate_fn translate) {struct cxl_root_decoder *cxlrd;
@@ -1840,7 +1823,6 @@ struct cxl_root_decoder *cxl_root_decoder_alloc(struct cxl_port *port,return ERR_PTR (rc); } - cxlrd->calc_hb = calc_hb; cxlrd->translate = translate;mutex_init (&cxlrd->range_lock);
drivers/cxl/cxl.h +0 -6
@@ -432,15 +432,12 @@ struct cxl_switch_decoder { };struct cxl_root_decoder ; -typedef struct cxl_dport *(*cxl_calc_hb_fn )(struct cxl_root_decoder *cxlrd, -int pos);typedef u64 (*cxl_translate_fn )(struct cxl_root_decoder *cxlrd,u64 hpa);/** * struct cxl_root_decoder - Static platform CXL address decoder * @res: host / parent resource for region allocations * @region_id: region id for next region provisioning event -* @calc_hb: which host bridge covers the n'th position by granularity * @translate: decoder specific address translation function * @platform_data: platform specific configuration data * @range_lock: sync region autodiscovery by address range
@@ -450,7 +447,6 @@ typedef u64 (*cxl_translate_fn)(struct cxl_root_decoder *cxlrd, u64 hpa);struct cxl_root_decoder {struct resource *res;atomic_t region_id; -cxl_calc_hb_fn calc_hb;cxl_translate_fn translate;void *platform_data;struct mutex range_lock;
@@ -776,9 +772,7 @@ bool is_switch_decoder(struct device *dev);bool is_endpoint_decoder (struct device *dev);struct cxl_root_decoder *cxl_root_decoder_alloc (struct cxl_port *port,unsigned int nr_targets, -cxl_calc_hb_fn calc_hb,cxl_translate_fn translate); -struct cxl_dport *cxl_hb_modulo (struct cxl_root_decoder *cxlrd,int pos);struct cxl_switch_decoder *cxl_switch_decoder_alloc (struct cxl_port *port,unsigned int nr_targets);int cxl_decoder_add (struct cxl_decoder *cxld,int *target_map);
Signature · 1 line
2.37.3
Re: [PATCH v2 1/4] cxl/core: Rename cxl_trace_hpa() to cxl_translate()
In reply to [email protected]Quoted 9 lines · alison.schofield@ wrote:
From: Alison Schofield <[email protected]>
Although
cxl_trace_hpa()is used to populate TRACE EVENTs with HPA addresses, the work it performs is a translation (dpa->hpa), not a trace. Rename it. Since this is the only translate work in CXL, drop the_hpasuffix in the rename.Suggested-by Dan Williams <[email protected]>Signed-off-by Alison Schofield <[email protected]>
LGTM
Re: [PATCH v2 2/4] cxl/acpi: Restore XOR'd position bits during address translation
In reply to [email protected]Quoted 8 lines · alison.schofield@ wrote:
From: Alison Schofield <[email protected]>
When a CXL region is created in a CXL Window (CFMWS) that uses XOR interleave arithmetic XOR maps are applied during the
HPA->DPAtranslation. The XOR function changes the interleave selector bit (aka position bit) in the HPA thereby varying which host bridge services an HPA. The purpose is to minimize hot spots thereby improving performance.
I think it is important to either detail how "interleave hot spots" arise, or just omit that description since it is ultimately not relevant to Linux. I.e. the "why XOR" question is irrelevant compared to "what happens to a end user with a kernel that does not comprehend XOR interleave maths".
Quoted 9 lines
When a device reports a DPA in events such as poison,
general_media, and dram, the driver translates that DPA back to an HPA. Presently, the CXL driver translation only considers the modulo position and will report the wrong HPA for XOR configured CFMWS's.Add a helper function that restores the XOR'd bits during
DPA->HPAaddress translation. Plumb a root decoder callback to the new helper when XOR interleave arithmetic is in use. For MODULO arithmetic, just let the callback be NULL - as in no extra work required.
Perhaps add a handful of sentences about the testing for this patch to make sure the maths are now correct compared to what was there before?
Quoted 41 lines
Fixes 28a3ae4ff66c ("cxl/trace: Add an HPA to cxl_poison trace events")Signed-off-by Alison Schofield <[email protected]>--- drivers/cxl/acpi.c | 48 ++++++++++++++++++++++++++++++++++++--- drivers/cxl/core/port.c | 5 +++- drivers/cxl/core/region.c | 5 ++++ drivers/cxl/cxl.h | 6 ++++- 4 files changed, 59 insertions(+), 5 deletions(-)drivers/cxl/acpi.c +23
-0@@ -74,6 +74,43 @@ static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos)return cxlrd->cxlsd.target[n]; } +static u64 cxl_xor_translate (struct cxl_root_decoder *cxlrd,u64 hpa) +{ +struct cxl_cxims_data *cximsd = cxlrd->platform_data; +int hbiw = cxlrd->cxlsd.nr_targets; +u64 val; +int pos; + +/* No xormaps for host bridge interleave ways of 1 or 3 */ +if (hbiw ==1 || hbiw ==3 ) +return hpa; + +/* +* For root decoders using xormaps (hbiw: 2,4,6,8,12,16) restore +* the position bit to its value before the xormap was applied at +* HPA->DPA translation. +* +* pos is the lowest set bit in an XORMAP +* val is the XORALLBITS(HPA & XORMAP) +* +* XORALLBITS: The CXL spec (3.1 Table 9-22) defines XORALLBITS +* as an operation that outputs a single bit by XORing all the +* bits in the input (hpa & xormap). Implement XORALLBITS using +* hweight64(). If the hamming weight is even the XOR of those
drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 109
Quoted 10 lines
drivers/cxl/acpi.c +10
-0+* bits results in 0, if odd the XOR result is 1. +*/ + +for (int i =0 ; i < cximsd->nr_maps; i++) { +if (!cximsd->xormaps[i]) +continue ; + pos =__ffs (cximsd->xormaps[i]); + val = (hweight64 (hpa & cximsd->xormaps[i]) &1 ); + hpa = (hpa & ~(1ULL << pos)) | (val << pos); + }
This looks correct to me, but so did the original broken implementation. I would feel better about a self test either integrated into tools/testing/cxl/, or documented / referenced in the CXL Device Driver Writer's Guide that gives confidence that "This is the way".
Alison Schofield
maintainerRe: [PATCH v2 2/4] cxl/acpi: Restore XOR'd position bits during address translation
Quoted 10 lines · Dan Williams
Quoted 8 lines · alison.schofield@ wrote:
From: Alison Schofield <[email protected]>
When a CXL region is created in a CXL Window (CFMWS) that uses XOR interleave arithmetic XOR maps are applied during the
HPA->DPAtranslation. The XOR function changes the interleave selector bit (aka position bit) in the HPA thereby varying which host bridge services an HPA. The purpose is to minimize hot spots thereby improving performance.
Quoted 5 lines
I think it is important to either detail how "interleave hot spots" arise, or just omit that description since it is ultimately not relevant to Linux. I.e. the "why XOR" question is irrelevant compared to "what happens to a end user with a kernel that does not comprehend XOR interleave maths".
I think I understand. Long ago we decided to add XOR Math support to the CXL Driver. The point now is to do it correctly not rejustify it's existence.
I'll chop 'The purpose...'
Quoted 6 lines
Quoted 6 lines
When a device reports a DPA in events such as poison,
general_media, and dram, the driver translates that DPA back to an HPA. Presently, the CXL driver translation only considers the modulo position and will report the wrong HPA for XOR configured CFMWS's.
Quoted 8 lines
Quoted 4 lines
Add a helper function that restores the XOR'd bits during
DPA->HPAaddress translation. Plumb a root decoder callback to the new helper when XOR interleave arithmetic is in use. For MODULO arithmetic, just let the callback be NULL - as in no extra work required.Perhaps add a handful of sentences about the testing for this patch to make sure the maths are now correct compared to what was there before?
There was nothing there before, but I think I know what you mean. More on that below...
Quoted 4 lines
Quoted 4 lines
Fixes 28a3ae4ff66c ("cxl/trace: Add an HPA to cxl_poison trace events")Signed-off-by Alison Schofield <[email protected]>---
snip
Quoted 5 lines
Quoted 5 lines
drivers/cxl/acpi.c +5
-0+ +for (int i =0 ; i < cximsd->nr_maps; i++) { +if (!cximsd->xormaps[i]) +continue ; + pos =__ffs (cximsd->xormaps[i]);
Quoted 8 lines
drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 109
Quoted 3 lines
drivers/cxl/acpi.c +3
-0+ val = (hweight64 (hpa & cximsd->xormaps[i]) &1 ); + hpa = (hpa & ~(1ULL << pos)) | (val << pos); + }This looks correct to me, but so did the original broken implementation. I would feel better about a self test either integrated into
tools/testing/cxl/, or documented / referenced in the CXL Device Driver Writer's Guide that gives confidence that "This is the way".
Let me see about adding a unit test to the cxl suite that exercises the translation path using data from the sample XOR translation tables that are currently being added to the CXL Driver Writer's Guide. That would allow the test to verify against a known set of DPA->HPA offsets for a given XOR region config.
-- Alison
Re: [PATCH v2 2/4] cxl/acpi: Restore XOR'd position bits during address translation
Quoted 5 lines · Alison Schofield
Let me see about adding a unit test to the cxl suite that exercises the translation path using data from the sample XOR translation tables that are currently being added to the CXL Driver Writer's Guide. That would allow the test to verify against a known set of
DPA->HPAoffsets for a given XOR region config.
Yeah, that sounds good. No need to worry about integrating that into cxl_test. It could just be a sample test program in cxl-cli that shows the algorithm and then can be referenced as "see, the kernel does the same thing as this test that passes with the sample data from the Driver Writer's Guide.
Re: [PATCH v2 1/4] cxl/core: Rename cxl_trace_hpa() to cxl_translate()
In reply to [email protected]Quoted 9 lines · On Wed, 8 May 2024 11:47:50 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
Although
cxl_trace_hpa()is used to populate TRACE EVENTs with HPA addresses, the work it performs is a translation (dpa->hpa), not a trace. Rename it. Since this is the only translate work in CXL, drop the_hpasuffix in the rename.Suggested-by Dan Williams <[email protected]>Signed-off-by Alison Schofield <[email protected]>
Maybe makes some sense to keep the direction in the name?
I can see for things like PPR we may need to go HPA to DPA as it'll be based on error counts that may be reported in HPA.
So Friday bikeshedding time.cxl_dpa_to_hpa()?
Quoted 65 lines
--- drivers/cxl/core/core.h | 4 ++-- drivers/cxl/core/mbox.c | 2 +- drivers/cxl/core/region.c | 2 +- drivers/cxl/core/trace.h | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-)drivers/cxl/core/core.h +2
-2@@ -28,12 +28,12 @@ int cxl_region_init(void);void cxl_region_exit (void );int cxl_get_poison_by_endpoint (struct cxl_port *port);struct cxl_region *cxl_dpa_to_region (const struct cxl_memdev *cxlmd,u64 dpa); -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa);#else static inline u64 -cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) +cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {return ULLONG_MAX ; }drivers/cxl/core/mbox.c +1
-1@@ -878,7 +878,7 @@ void cxl_event_trace_record(const struct cxl_memdev *cxlmd, dpa =le64_to_cpu (evt->common.phys_addr) &CXL_DPA_MASK ; cxlr =cxl_dpa_to_region (cxlmd, dpa);if (cxlr) - hpa =cxl_trace_hpa (cxlr, cxlmd, dpa); + hpa =cxl_translate (cxlr, cxlmd, dpa);if (event_type ==CXL_CPER_EVENT_GEN_MEDIA ) trace_cxl_general_media(cxlmd, type, cxlr, hpa,drivers/cxl/core/region.c +1
-1@@ -2797,7 +2797,7 @@ static u64 cxl_dpa_to_hpa(u64 dpa, struct cxl_region *cxlr,return hpa; } -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {struct cxl_region_params *p = &cxlr->params;drivers/cxl/core/trace.h +1
-1@@ -704,7 +704,7 @@ TRACE_EVENT(cxl_poison,if (cxlr) {__assign_str (region,dev_name (&cxlr->dev));memcpy (__entry->uuid, &cxlr->params.uuid,16 ); - __entry->hpa =cxl_trace_hpa (cxlr, cxlmd, + __entry->hpa =cxl_translate (cxlr, cxlmd, __entry->dpa); }else {__assign_str (region,"" );
Re: [PATCH v2 2/4] cxl/acpi: Restore XOR'd position bits during address translation
In reply to [email protected]Quoted 14 lines · On Wed, 8 May 2024 11:47:51 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
When a CXL region is created in a CXL Window (CFMWS) that uses XOR interleave arithmetic XOR maps are applied during the
HPA->DPAtranslation. The XOR function changes the interleave selector bit (aka position bit) in the HPA thereby varying which host bridge services an HPA. The purpose is to minimize hot spots thereby improving performance.When a device reports a DPA in events such as poison,
general_media, and dram, the driver translates that DPA back to an HPA. Presently, the CXL driver translation only considers the modulo position and will report the wrong HPA for XOR configured CFMWS's.
Quoted 7 lines
Add a helper function that restores the XOR'd bits during
DPA->HPAaddress translation. Plumb a root decoder callback to the new helper when XOR interleave arithmetic is in use. For MODULO arithmetic, just let the callback be NULL - as in no extra work required.Fixes 28a3ae4ff66c ("cxl/trace: Add an HPA to cxl_poison trace events")Signed-off-by Alison Schofield <[email protected]>
Trivial comment inline. Agree entirely that some tests would be good.
I ran through a few trivial cases on a bit of paper and it looks to me like it works but that hardly counts as testing :)
Quoted 35 lines
--- drivers/cxl/acpi.c | 48 ++++++++++++++++++++++++++++++++++++--- drivers/cxl/core/port.c | 5 +++- drivers/cxl/core/region.c | 5 ++++ drivers/cxl/cxl.h | 6 ++++- 4 files changed, 59 insertions(+), 5 deletions(-)drivers/cxl/acpi.c +20
-0@@ -74,6 +74,43 @@ static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos)return cxlrd->cxlsd.target[n]; } +static u64 cxl_xor_translate (struct cxl_root_decoder *cxlrd,u64 hpa) +{ +struct cxl_cxims_data *cximsd = cxlrd->platform_data; +int hbiw = cxlrd->cxlsd.nr_targets; +u64 val; +int pos; + +/* No xormaps for host bridge interleave ways of 1 or 3 */ +if (hbiw ==1 || hbiw ==3 ) +return hpa; + +/* +* For root decoders using xormaps (hbiw: 2,4,6,8,12,16) restore +* the position bit to its value before the xormap was applied at +* HPA->DPA translation. +* +* pos is the lowest set bit in an XORMAP +* val is the XORALLBITS(HPA & XORMAP) +* +* XORALLBITS: The CXL spec (3.1 Table 9-22) defines XORALLBITS
drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 106
Quoted 10 lines
drivers/cxl/acpi.c +10
-0+* as an operation that outputs a single bit by XORing all the +* bits in the input (hpa & xormap). Implement XORALLBITS using +* hweight64(). If the hamming weight is even the XOR of those +* bits results in 0, if odd the XOR result is 1. +*/ + +for (int i =0 ; i < cximsd->nr_maps; i++) { +if (!cximsd->xormaps[i]) +continue ; + pos =__ffs (cximsd->xormaps[i]);
At the moment the comment on XORALLBITS isn't associated with this code very well. I'd factor it out as cxl_xorallbits() mostly so you can stick the comment next to the bit that does the work. Or maybe a #define XORALLBITS(hpa, xormap) is good enough if you move it up under the comment.
Quoted 7 lines
drivers/cxl/acpi.c +7
-0+ val = (hweight64 (hpa & cximsd->xormaps[i]) &1 ); + hpa = (hpa & ~(1ULL << pos)) | (val << pos); + } + +return hpa; +} +
Re: [PATCH v2 3/4] cxl/region: Verify target positions using the ordered target list
In reply to [email protected]Quoted 13 lines · On Wed, 8 May 2024 11:47:52 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
When a root decoder is configured the interleave target list is read from the BIOS populated CFMWS structure. Per the CXL spec 3.1 Table 9-22 the target list is in interleave order. The CXL driver populates its decoder target list in the same order and stores it in 'struct
cxl_switch_decoder' field "@target: active ordered target list in current decoder configuration"Given the promise of an ordered list, the driver can stop duplicating the work of BIOS and simply check target positions against the ordered list during region configuration.
Quoted 8 lines
The simplified check against the ordered list is presented here. A follow-on patch will remove the unused code.
For Modulo arithmetic this is not a fix, only a simplification. For XOR arithmetic this is a fix for HB IW of 3,6,12.
Fixes f9db85bfec0d ("cxl/acpi: Support CXL XOR Interleave Math (CXIMS)")Signed-off-by Alison Schofield <[email protected]>
Re: [PATCH v2 4/4] cxl: Remove defunct code calculating host bridge target positions
In reply to [email protected]Quoted 7 lines · On Wed, 8 May 2024 11:47:53 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
The CXL Spec 3.1 Table 9-22 requires that the BIOS populate the CFMWS target list in interleave target order. This means that the calculations the CXL driver added to determine positions when XOR math is in use, along with the entire XOR vs Modulo call back setup is not needed.
Quoted 6 lines
A prior patch added a common method to verify positions.
Remove the now unused code related to the
cxl_calc_hb_fn.Signed-off-by Alison Schofield <[email protected]>Reviewed-by Dan Williams <[email protected]>
FWIW given this just removes unused code.
Alison Schofield
maintainerRe: [PATCH v2 1/4] cxl/core: Rename cxl_trace_hpa() to cxl_translate()
In reply to Jonathan CameronQuoted 13 lines · Jonathan Cameron
Quoted 9 lines · On Wed, 8 May 2024 11:47:50 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
Although
cxl_trace_hpa()is used to populate TRACE EVENTs with HPA addresses, the work it performs is a translation (dpa->hpa), not a trace. Rename it. Since this is the only translate work in CXL, drop the_hpasuffix in the rename.Suggested-by Dan Williams <[email protected]>Signed-off-by Alison Schofield <[email protected]>
Quoted 7 lines
Maybe makes some sense to keep the direction in the name?
I can see for things like PPR we may need to go HPA to DPA as it'll be based on error counts that may be reported in HPA.
So Friday bikeshedding time.
cxl_dpa_to_hpa()?
Thanks for the review Jonathan!
Let's sync on the possible dpa->hpa->dpa use cases and then I'll weigh in on your bikeshedding.
Agree we need HPA->DPA support for userspace. User needs ability to find which device maps an HPA.
wrt PPR, do you think that is an in kernel need?
If the PPR maintenance was done in response to a CXL event, the DPA is already in the event trace along with the HPA so I wouldn't expect that a lookup is needed, at least not within the kernel.
Your explicit cxl_dpa_to_hpa() name is good.
--Alison
Quoted 68 lines
Quoted 67 lines
--- drivers/cxl/core/core.h | 4 ++-- drivers/cxl/core/mbox.c | 2 +- drivers/cxl/core/region.c | 2 +- drivers/cxl/core/trace.h | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-)drivers/cxl/core/core.h +2
-2@@ -28,12 +28,12 @@ int cxl_region_init(void);void cxl_region_exit (void );int cxl_get_poison_by_endpoint (struct cxl_port *port);struct cxl_region *cxl_dpa_to_region (const struct cxl_memdev *cxlmd,u64 dpa); -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa);#else static inline u64 -cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) +cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {return ULLONG_MAX ; }drivers/cxl/core/mbox.c +1
-1@@ -878,7 +878,7 @@ void cxl_event_trace_record(const struct cxl_memdev *cxlmd, dpa =le64_to_cpu (evt->common.phys_addr) &CXL_DPA_MASK ; cxlr =cxl_dpa_to_region (cxlmd, dpa);if (cxlr) - hpa =cxl_trace_hpa (cxlr, cxlmd, dpa); + hpa =cxl_translate (cxlr, cxlmd, dpa);if (event_type ==CXL_CPER_EVENT_GEN_MEDIA ) trace_cxl_general_media(cxlmd, type, cxlr, hpa,drivers/cxl/core/region.c +1
-1@@ -2797,7 +2797,7 @@ static u64 cxl_dpa_to_hpa(u64 dpa, struct cxl_region *cxlr,return hpa; } -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {struct cxl_region_params *p = &cxlr->params;drivers/cxl/core/trace.h +1
-1@@ -704,7 +704,7 @@ TRACE_EVENT(cxl_poison,if (cxlr) {__assign_str (region,dev_name (&cxlr->dev));memcpy (__entry->uuid, &cxlr->params.uuid,16 ); - __entry->hpa =cxl_trace_hpa (cxlr, cxlmd, + __entry->hpa =cxl_translate (cxlr, cxlmd, __entry->dpa); }else {__assign_str (region,"" );
Re: [PATCH v2 1/4] cxl/core: Rename cxl_trace_hpa() to cxl_translate()
Quoted 27 lines · On Fri, 7 Jun 2024 10:45:42 -0700 Alison Schofield…
Quoted 13 lines · Jonathan Cameron
Quoted 9 lines · On Wed, 8 May 2024 11:47:50 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
Although
cxl_trace_hpa()is used to populate TRACE EVENTs with HPA addresses, the work it performs is a translation (dpa->hpa), not a trace. Rename it. Since this is the only translate work in CXL, drop the_hpasuffix in the rename.Suggested-by Dan Williams <[email protected]>Signed-off-by Alison Schofield <[email protected]>Quoted 7 lines
Maybe makes some sense to keep the direction in the name?
I can see for things like PPR we may need to go HPA to DPA as it'll be based on error counts that may be reported in HPA.
So Friday bikeshedding time.
cxl_dpa_to_hpa()?Thanks for the review Jonathan!
Let's sync on the possible
dpa->hpa->dpause cases and then I'll weigh in on your bikeshedding.
Quoted 7 lines
Agree we need
HPA->DPAsupport for userspace. User needs ability to find which device maps an HPA.wrt PPR, do you think that is an in kernel need?
If the PPR maintenance was done in response to a CXL event, the DPA is already in the event trace along with the HPA so I wouldn't expect that a lookup is needed, at least not within the kernel.
Might be a synchronous poison event that ends up triggering thing. And for that we'll be lucky to get any useful info beyond an HPA.
I've only started thinking about this in the last few days, so not 100% sure yet.
Given PPR can I think interrupt data flow we need to be very careful to mediate any attempt so to enable it at all will need kernel support. Whether the interface takes a DPA or an HPA is an open question. We'll see!
Jonathan
Quoted 76 lines
Your explicit
cxl_dpa_to_hpa()name is good.--Alison
Quoted 68 lines
Quoted 67 lines
--- drivers/cxl/core/core.h | 4 ++-- drivers/cxl/core/mbox.c | 2 +- drivers/cxl/core/region.c | 2 +- drivers/cxl/core/trace.h | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-)drivers/cxl/core/core.h +2
-2@@ -28,12 +28,12 @@ int cxl_region_init(void);void cxl_region_exit (void );int cxl_get_poison_by_endpoint (struct cxl_port *port);struct cxl_region *cxl_dpa_to_region (const struct cxl_memdev *cxlmd,u64 dpa); -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa);#else static inline u64 -cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) +cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {return ULLONG_MAX ; }drivers/cxl/core/mbox.c +1
-1@@ -878,7 +878,7 @@ void cxl_event_trace_record(const struct cxl_memdev *cxlmd, dpa =le64_to_cpu (evt->common.phys_addr) &CXL_DPA_MASK ; cxlr =cxl_dpa_to_region (cxlmd, dpa);if (cxlr) - hpa =cxl_trace_hpa (cxlr, cxlmd, dpa); + hpa =cxl_translate (cxlr, cxlmd, dpa);if (event_type ==CXL_CPER_EVENT_GEN_MEDIA ) trace_cxl_general_media(cxlmd, type, cxlr, hpa,drivers/cxl/core/region.c +1
-1@@ -2797,7 +2797,7 @@ static u64 cxl_dpa_to_hpa(u64 dpa, struct cxl_region *cxlr,return hpa; } -u64 cxl_trace_hpa (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd, +u64 cxl_translate (struct cxl_region *cxlr,const struct cxl_memdev *cxlmd,u64 dpa) {struct cxl_region_params *p = &cxlr->params;drivers/cxl/core/trace.h +1
-1@@ -704,7 +704,7 @@ TRACE_EVENT(cxl_poison,if (cxlr) {__assign_str (region,dev_name (&cxlr->dev));memcpy (__entry->uuid, &cxlr->params.uuid,16 ); - __entry->hpa =cxl_trace_hpa (cxlr, cxlmd, + __entry->hpa =cxl_translate (cxlr, cxlmd, __entry->dpa); }else {__assign_str (region,"" );
Alison Schofield
maintainerRe: [PATCH v2 2/4] cxl/acpi: Restore XOR'd position bits during address translation
In reply to Jonathan CameronQuoted 22 lines · Jonathan Cameron
Quoted 19 lines · On Wed, 8 May 2024 11:47:51 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
When a CXL region is created in a CXL Window (CFMWS) that uses XOR interleave arithmetic XOR maps are applied during the
HPA->DPAtranslation. The XOR function changes the interleave selector bit (aka position bit) in the HPA thereby varying which host bridge services an HPA. The purpose is to minimize hot spots thereby improving performance.When a device reports a DPA in events such as poison,
general_media, and dram, the driver translates that DPA back to an HPA. Presently, the CXL driver translation only considers the modulo position and will report the wrong HPA for XOR configured CFMWS's.Add a helper function that restores the XOR'd bits during
DPA->HPAaddress translation. Plumb a root decoder callback to the new helper when XOR interleave arithmetic is in use. For MODULO arithmetic, just let the callback be NULL - as in no extra work required.
Quoted 5 lines
Quoted 2 lines
Fixes 28a3ae4ff66c ("cxl/trace: Add an HPA to cxl_poison trace events")Signed-off-by Alison Schofield <[email protected]>Trivial comment inline. Agree entirely that some tests would be good.
I ran through a few trivial cases on a bit of paper and it looks to me like it works but that hardly counts as testing :)
Thanks for the review and for doing some calcs.
I've become very adept at working these out with paper/pencil, that hop to C implementation is the challenge ;)
Quoted 2 lines
Reviewed-by Jonathan Cameron <[email protected]>
Hmm...well, let me know if I can keep that after you read below...
Quoted 53 lines
Quoted 36 lines
--- drivers/cxl/acpi.c | 48 ++++++++++++++++++++++++++++++++++++--- drivers/cxl/core/port.c | 5 +++- drivers/cxl/core/region.c | 5 ++++ drivers/cxl/cxl.h | 6 ++++- 4 files changed, 59 insertions(+), 5 deletions(-)drivers/cxl/acpi.c +20
-0@@ -74,6 +74,43 @@ static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos)return cxlrd->cxlsd.target[n]; } +static u64 cxl_xor_translate (struct cxl_root_decoder *cxlrd,u64 hpa) +{ +struct cxl_cxims_data *cximsd = cxlrd->platform_data; +int hbiw = cxlrd->cxlsd.nr_targets; +u64 val; +int pos; + +/* No xormaps for host bridge interleave ways of 1 or 3 */ +if (hbiw ==1 || hbiw ==3 ) +return hpa; + +/* +* For root decoders using xormaps (hbiw: 2,4,6,8,12,16) restore +* the position bit to its value before the xormap was applied at +* HPA->DPA translation. +* +* pos is the lowest set bit in an XORMAP +* val is the XORALLBITS(HPA & XORMAP) +* +* XORALLBITS: The CXL spec (3.1 Table 9-22) defines XORALLBITS drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 106
Quoted 10 lines
drivers/cxl/acpi.c +10
-0+* as an operation that outputs a single bit by XORing all the +* bits in the input (hpa & xormap). Implement XORALLBITS using +* hweight64(). If the hamming weight is even the XOR of those +* bits results in 0, if odd the XOR result is 1. +*/ + +for (int i =0 ; i < cximsd->nr_maps; i++) { +if (!cximsd->xormaps[i]) +continue ; + pos =__ffs (cximsd->xormaps[i]);At the moment the comment on XORALLBITS isn't associated with this code very well. I'd factor it out as
cxl_xorallbits()mostly so you can stick the comment next to the bit that does the work. Or maybe a #defineXORALLBITS(hpa, xormap)is good enough if you move it up under the comment.
Quoted 7 lines
drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 113
Quoted 7 lines
drivers/cxl/acpi.c +7
-0+ val = (hweight64 (hpa & cximsd->xormaps[i]) &1 ); + hpa = (hpa & ~(1ULL << pos)) | (val << pos); + } + +return hpa; +} +
You haven't convinced me that readers will not be able to associate the block comment directly above the for-loop with the work inside the for-loop. Especially since this is a 25 line function with a single focus.
I intentionally didn't insert line-by-line commentary in the for loop, but rather told the story in the comment and then just did it.
Maybe repeating 'val' here, wraps up the comment better:
- * bits results in 0, if odd the XOR result is 1.
+ * bits results in val==0, if odd the XOR results in val==1.
-- Alison
Quoted 2 lines
Re: [PATCH v2 2/4] cxl/acpi: Restore XOR'd position bits during address translation
Quoted 31 lines · On Fri, 7 Jun 2024 11:20:25 -0700 Alison Schofield…
Quoted 22 lines · Jonathan Cameron
Quoted 19 lines · On Wed, 8 May 2024 11:47:51 -0700 [email protected] wrote:
From: Alison Schofield <[email protected]>
When a CXL region is created in a CXL Window (CFMWS) that uses XOR interleave arithmetic XOR maps are applied during the
HPA->DPAtranslation. The XOR function changes the interleave selector bit (aka position bit) in the HPA thereby varying which host bridge services an HPA. The purpose is to minimize hot spots thereby improving performance.When a device reports a DPA in events such as poison,
general_media, and dram, the driver translates that DPA back to an HPA. Presently, the CXL driver translation only considers the modulo position and will report the wrong HPA for XOR configured CFMWS's.Add a helper function that restores the XOR'd bits during
DPA->HPAaddress translation. Plumb a root decoder callback to the new helper when XOR interleave arithmetic is in use. For MODULO arithmetic, just let the callback be NULL - as in no extra work required.Quoted 5 lines
Quoted 2 lines
Fixes 28a3ae4ff66c ("cxl/trace: Add an HPA to cxl_poison trace events")Signed-off-by Alison Schofield <[email protected]>Trivial comment inline. Agree entirely that some tests would be good.
I ran through a few trivial cases on a bit of paper and it looks to me like it works but that hardly counts as testing :)Thanks for the review and for doing some calcs.
Quoted 7 lines
I've become very adept at working these out with paper/pencil, that hop to C implementation is the challenge ;)
Quoted 2 lines
Reviewed-by Jonathan Cameron <[email protected]>Hmm...well, let me know if I can keep that after you read below...
Meh. I don't care that much. Just made me thing a tiny bit more
than I like to ;) Fine with or without the change you suggest.
Jonathan
Quoted 89 lines
Quoted 54 lines
Quoted 37 lines
--- drivers/cxl/acpi.c | 48 ++++++++++++++++++++++++++++++++++++--- drivers/cxl/core/port.c | 5 +++- drivers/cxl/core/region.c | 5 ++++ drivers/cxl/cxl.h | 6 ++++- 4 files changed, 59 insertions(+), 5 deletions(-)drivers/cxl/acpi.c +20
-0@@ -74,6 +74,43 @@ static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos)return cxlrd->cxlsd.target[n]; } +static u64 cxl_xor_translate (struct cxl_root_decoder *cxlrd,u64 hpa) +{ +struct cxl_cxims_data *cximsd = cxlrd->platform_data; +int hbiw = cxlrd->cxlsd.nr_targets; +u64 val; +int pos; + +/* No xormaps for host bridge interleave ways of 1 or 3 */ +if (hbiw ==1 || hbiw ==3 ) +return hpa; + +/* +* For root decoders using xormaps (hbiw: 2,4,6,8,12,16) restore +* the position bit to its value before the xormap was applied at +* HPA->DPA translation. +* +* pos is the lowest set bit in an XORMAP +* val is the XORALLBITS(HPA & XORMAP) +* +* XORALLBITS: The CXL spec (3.1 Table 9-22) defines XORALLBITS drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 106
Quoted 10 lines
drivers/cxl/acpi.c +10
-0+* as an operation that outputs a single bit by XORing all the +* bits in the input (hpa & xormap). Implement XORALLBITS using +* hweight64(). If the hamming weight is even the XOR of those +* bits results in 0, if odd the XOR result is 1. +*/ + +for (int i =0 ; i < cximsd->nr_maps; i++) { +if (!cximsd->xormaps[i]) +continue ; + pos =__ffs (cximsd->xormaps[i]);At the moment the comment on XORALLBITS isn't associated with this code very well. I'd factor it out as
cxl_xorallbits()mostly so you can stick the comment next to the bit that does the work. Or maybe a #defineXORALLBITS(hpa, xormap)is good enough if you move it up under the comment.Quoted 7 lines
drivers/cxl/acpi.c static struct cxl_dport *cxl_hb_xor(struct cxl_root_decoder *cxlrd, int pos) · line 113
Quoted 7 lines
drivers/cxl/acpi.c +7
-0+ val = (hweight64 (hpa & cximsd->xormaps[i]) &1 ); + hpa = (hpa & ~(1ULL << pos)) | (val << pos); + } + +return hpa; +} +You haven't convinced me that readers will not be able to associate the block comment directly above the for-loop with the work inside the for-loop. Especially since this is a 25 line function with a single focus.
I intentionally didn't insert line-by-line commentary in the for loop, but rather told the story in the comment and then just did it.
Maybe repeating 'val' here, wraps up the comment better:
(quoted diff) +1
-2- * bits results in 0, if odd the XOR result is 1. + * bits results in val==0, if odd the XOR results in val==1. -- AlisonQuoted 11 lines
Re: [PATCH v2 3/4] cxl/region: Verify target positions using the ordered target list
In reply to [email protected]Quoted 13 lines · alison.schofield@ wrote:
From: Alison Schofield <[email protected]>
When a root decoder is configured the interleave target list is read from the BIOS populated CFMWS structure. Per the CXL spec 3.1 Table 9-22 the target list is in interleave order. The CXL driver populates its decoder target list in the same order and stores it in 'struct
cxl_switch_decoder' field "@target: active ordered target list in current decoder configuration"Given the promise of an ordered list, the driver can stop duplicating the work of BIOS and simply check target positions against the ordered list during region configuration.
Quoted 8 lines
The simplified check against the ordered list is presented here. A follow-on patch will remove the unused code.
For Modulo arithmetic this is not a fix, only a simplification. For XOR arithmetic this is a fix for HB IW of 3,6,12.
Fixes f9db85bfec0d ("cxl/acpi: Support CXL XOR Interleave Math (CXIMS)")Signed-off-by Alison Schofield <[email protected]>
LGTM