Linux-ARM-Kernel Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Jinjie Ruan <ruanjinjie@huawei.com>
To: Ben Horgan <ben.horgan@arm.com>, <james.morse@arm.com>,
	<reinette.chatre@intel.com>, <fenghuay@nvidia.com>,
	<tony.luck@intel.com>, <Dave.Martin@arm.com>,
	<babu.moger@amd.com>, <will@kernel.org>, <jic23@kernel.org>,
	<gshan@redhat.com>, <bp@alien8.de>, <x86@kernel.org>,
	<linux-arm-kernel@lists.infradead.org>,
	<linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 1/2] arm_mpam: resctrl: Use the domain's counter ID
Date: Tue, 17 Nov 2026 19:48:23 +0800	[thread overview]
Message-ID: <f5cc83eb-7c67-4db0-add4-08e893640f92@huawei.com> (raw)
In-Reply-To: <a7dc44ee-cf2c-4e2a-9953-4f8ffeffda60@arm.com>

Hi Ben,

Thanks for the review.

在 2026/11/17 19:27, Ben Horgan 写道:
> Hi Jinjie,
> 
> On 13/11/2026 08:35, tizi4590 wrote:
>> From: Jinjie Ruan <ruanjinjie@huawei.com>
>>
>> resctrl assigns MBM counters per domain: mbm_L3_assignments takes a
>> domain ID, a group can have a counter in one domain and none in
>> another, and resctrl_arch_cntr_read() and resctrl_arch_reset_cntr()
>> are given the counter to use.
>>
>> The driver ignores that counter and looks the monitor up in
>> mbwu_idx_to_mon[], indexed by CLOSID and RMID.
>> resctrl_arch_config_cntr() fills that array without looking at the
>> domain, so each group has one monitor for the whole resource.
>> Whichever assignment is written last selects the monitor the group
>> reads in every domain.
> 
> For me, this scenario where we read the wrong counter to be the most
> serious bug in the MPAM driver I currently know about. Thanks for
> addressing it.

It is a great honor to be able to participate in solving this issue.

> 
>>
>> For example, with the counter ID each group uses for mbm_total_bytes
>> on each domain and "_" for none:
>>
>>                                 domain 0         domain 1
>>   mount, mkdir g1               default=0 g1=1   default=0 g1=1
>>   unassign both on domain 1     default=0 g1=1   default=_ g1=_
>>   assign g1 on domain 1         default=0 g1=1   default=_ g1=0
>>
>> g1 is assigned the lowest free counter on domain 1, counter 0. That
>> assignment is the last one written for g1, so it selects the monitor
>> g1 reads in every domain. On domain 0, g1 uses counter 1, but reads
>> the monitor of counter 0:
>>
>>                       counter whose monitor is read on domain 0
>>                       default      g1
>>   expected            0            1
>>   actual              0            0
>>
>> That is the default group's monitor.
>>
>> Translate the counter ID with assigned_counters[], which maps a counter
>> ID to a monitor of the class and is valid in every domain, 
> 
> Ok, sounds reasonable.
> 
> and drop
>> mbwu_idx_to_mon[] with the lookups in __read_mon() and __reset_mon().

Will drop this code description.

> 
> mbwu_idx_to_mon[] does become unused but if we want to ever support the
> default mbm_assign_mode then it may become used. A little more on this
> below but this can be considered separately from the bug fix.
> 
>> With nothing left to record, resctrl_arch_config_cntr() only resets the
>> monitors of a counter when it is assigned.
> 
> Isn't this a separate optimization?

Not really, just a description of the current state after the fix
(removing mbwu_idx_to_mon())

> 
>>
>> Nothing consumes USE_PRE_ALLOCATED, the marker that triggered those
>> lookups, once they are gone. resctrl only allocates a monitor context
>> for an MBM event when resctrl_arch_mbm_cntr_assign_enabled() reports
>> false, and this driver always reports true for the L3, so the MBM
>> events are always read as assigned counters. Remove the marker, and
>> the MBM cases of resctrl_arch_mon_ctx_alloc_no_wait() that returned it.
> 
> True but this is a separate, benign, issue with the code.
> resctrl_arch_mon_ctx_alloc() is only called for mbm counters when
> mbm_assign_mode is default rather than mbm_event. In MPAM, we currently
> only support the mbm_event mode, assigning counters. Crossing threads,
> but you had an rfc for supporting the software controller on MPAM wher
> using the default mbm_assign_mode can be considered. Regardless, please
> can you keep the changes in this patch only to those addressing the
> particular bug and other changes can be considered as separate patches.
> 
>>
>> Fixes: a5ff6ef2d73e ("arm_mpam: resctrl: Add resctrl_arch_cntr_read() & resctrl_arch_reset_cntr()")
>> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
>> Cc: <stable@vger.kernel.org>
>> ---
>>  drivers/resctrl/mpam_internal.h |  15 ---
>>  drivers/resctrl/mpam_resctrl.c  | 195 ++++++++++++--------------------
>>  2 files changed, 71 insertions(+), 139 deletions(-)
>>
>> diff --git a/drivers/resctrl/mpam_internal.h b/drivers/resctrl/mpam_internal.h
>> index 1f86cc158100..e1348d958d05 100644
>> --- a/drivers/resctrl/mpam_internal.h
>> +++ b/drivers/resctrl/mpam_internal.h
>> @@ -30,14 +30,6 @@ struct platform_device;
>>  #define PACKED_FOR_KUNIT
>>  #endif
>>  
>> -/*
>> - * This 'mon' values must not alias an actual monitor, so must be larger than
>> - * U16_MAX, but not be confused with an errno value, so smaller than
>> - * (u32)-SZ_4K.
>> - * USE_PRE_ALLOCATED is used to avoid confusion with an actual monitor.
>> - */
>> -#define USE_PRE_ALLOCATED	(U16_MAX + 1)
>> -
>>  static inline bool mpam_is_enabled(void)
>>  {
>>  	return static_branch_likely(&mpam_enabled);
>> @@ -309,10 +301,6 @@ enum mon_filter_options {
>>  };
>>  
>>  struct mon_cfg {
>> -	/*
>> -	 * mon must be large enough to hold out of range values like
>> -	 * USE_PRE_ALLOCATED
>> -	 */
>>  	u32			mon;
>>  	u8			pmg;
>>  	bool			match_pmg;
>> @@ -455,9 +443,6 @@ struct mpam_resctrl_res {
>>  struct mpam_resctrl_mon {
>>  	struct mpam_class	*class;
>>  
>> -	/* Array of allocated MBWU monitors, indexed by (closid, rmid). */
>> -	int			*mbwu_idx_to_mon;
>> -
>>  	/* Array of assigned MBWU monitors, indexed by resctrl's cntr_id. */
>>  	int			*assigned_counters;
>>  };
>> diff --git a/drivers/resctrl/mpam_resctrl.c b/drivers/resctrl/mpam_resctrl.c
>> index b28c5c04a87c..05c1bf97c43d 100644
>> --- a/drivers/resctrl/mpam_resctrl.c
>> +++ b/drivers/resctrl/mpam_resctrl.c
>> @@ -397,20 +397,13 @@ static int resctrl_arch_mon_ctx_alloc_no_wait(enum resctrl_event_id evtid)
>>  {
>>  	struct mpam_resctrl_mon *mon = &mpam_resctrl_counters[evtid];
>>  
>> -	if (!mpam_is_enabled())
>> -		return -EINVAL;
>> -
>> -	if (!mon->class)
>> +	if (!mpam_is_enabled() || !mon->class)
>>  		return -EINVAL;
> 
> I don't see any reason for this refactoring.


Here, the same error return is combined just for cleanup.

> 
>>  
>> -	switch (evtid) {
>> -	case QOS_L3_OCCUP_EVENT_ID:
>> +	if (evtid == QOS_L3_OCCUP_EVENT_ID) {
>>  		/* With CDP, one monitor gets used for both code/data reads */
>>  		return mpam_alloc_csu_mon(mon->class);
>> -	case QOS_L3_MBM_LOCAL_EVENT_ID:
>> -	case QOS_L3_MBM_TOTAL_EVENT_ID:
>> -		return USE_PRE_ALLOCATED;
>> -	default:
>> +	} else {
>>  		return -EOPNOTSUPP;
>>  	}
>>  }
>> @@ -464,9 +457,15 @@ void resctrl_arch_mon_ctx_free(struct rdt_resource *r,
>>  	resctrl_arch_mon_ctx_free_no_wait(evtid, mon_idx);
>>  }
>>  
>> -static int __read_mon(struct mpam_resctrl_mon *mon, struct mpam_component *mon_comp,
>> -		      enum mpam_device_features mon_type,
>> -		      int mon_idx,
>> +/* The monitor resctrl's counter @cntr_id uses for @cdp_type. */
>> +static int mon_for_cntr(struct mpam_resctrl_mon *mon, int cntr_id,
>> +			enum resctrl_conf_type cdp_type)
>> +{
>> +	return mon->assigned_counters[resctrl_get_config_index(cntr_id, cdp_type)];
>> +}
>> +
>> +static int __read_mon(struct mpam_component *mon_comp,
>> +		      enum mpam_device_features mon_type, int mon_idx,
>>  		      enum resctrl_conf_type cdp_type, u32 closid, u32 rmid, u64 *val)
>>  {
>>  	struct mon_cfg cfg;
>> @@ -477,14 +476,6 @@ static int __read_mon(struct mpam_resctrl_mon *mon, struct mpam_component *mon_c
>>  	/* Shift closid to account for CDP */
>>  	closid = resctrl_get_config_index(closid, cdp_type);
>>  
>> -	if (mon_idx == USE_PRE_ALLOCATED) {
>> -		int mbwu_idx = resctrl_arch_rmid_idx_encode(closid, rmid);
>> -
>> -		mon_idx = mon->mbwu_idx_to_mon[mbwu_idx];
>> -		if (mon_idx == -1)
>> -			return -ENOENT;
>> -	}
>> -
>>  	if (irqs_disabled()) {
>>  		/* Check if we can access this domain without an IPI */
>>  		return -EIO;
>> @@ -500,32 +491,6 @@ static int __read_mon(struct mpam_resctrl_mon *mon, struct mpam_component *mon_c
>>  	return mpam_msmon_read(mon_comp, &cfg, mon_type, val);
>>  }
>>  
>> -static int read_mon_cdp_safe(struct mpam_resctrl_mon *mon, struct mpam_component *mon_comp,
>> -			     enum mpam_device_features mon_type,
>> -			     int mon_idx, u32 closid, u32 rmid, u64 *val)
>> -{
>> -	if (cdp_enabled) {
>> -		u64 code_val = 0, data_val = 0;
>> -		int err;
>> -
>> -		err = __read_mon(mon, mon_comp, mon_type, mon_idx,
>> -				 CDP_CODE, closid, rmid, &code_val);
>> -		if (err)
>> -			return err;
>> -
>> -		err = __read_mon(mon, mon_comp, mon_type, mon_idx,
>> -				 CDP_DATA, closid, rmid, &data_val);
>> -		if (err)
>> -			return err;
>> -
>> -		*val = code_val + data_val;
>> -		return 0;
>> -	}
>> -
>> -	return __read_mon(mon, mon_comp, mon_type, mon_idx,
>> -			  CDP_NONE, closid, rmid, val);
>> -}
>> -
>>  /* MBWU when not in ABMC mode (not supported), and CSU counters. */
>>  int resctrl_arch_rmid_read(struct rdt_resource *r, struct rdt_domain_hdr *hdr,
>>  			   u32 closid, u32 rmid, enum resctrl_event_id eventid,
>> @@ -553,13 +518,31 @@ int resctrl_arch_rmid_read(struct rdt_resource *r, struct rdt_domain_hdr *hdr,
>>  
>>  	mon_type = mpam_feat_msmon_csu;
>>  
>> -	return read_mon_cdp_safe(mon, mon_comp, mon_type, mon_idx,
>> -				 closid, rmid, val);
>> +	if (cdp_enabled) {
>> +		u64 code_val = 0, data_val = 0;
>> +		int err;
>> +
>> +		err = __read_mon(mon_comp, mon_type, mon_idx,
>> +				 CDP_CODE, closid, rmid, &code_val);
>> +		if (err)
>> +			return err;
>> +
>> +		err = __read_mon(mon_comp, mon_type, mon_idx,
>> +				 CDP_DATA, closid, rmid, &data_val);
>> +		if (err)
>> +			return err;
>> +
>> +		*val = code_val + data_val;
>> +		return 0;
>> +	}
>> +
>> +	return __read_mon(mon_comp, mon_type, mon_idx,
>> +			  CDP_NONE, closid, rmid, val);
> 
> 
> What am I missing? Aren't you just copying the contents of
> read_mon_cdp_safe()?

After this patch, read_mon_cdp_safe no longer has a second caller, so
remove it.

> 
> Thanks,
> 
> Ben
> 
>>  }
>>  
>>  /* MBWU counters when in ABMC mode */
>>  int resctrl_arch_cntr_read(struct rdt_resource *r, struct rdt_l3_mon_domain *d,
>> -			   u32 closid, u32 rmid, int mon_idx,
>> +			   u32 closid, u32 rmid, int cntr_id,
>>  			   enum resctrl_event_id eventid, u64 *val)
>>  {
>>  	struct mpam_resctrl_mon *mon = &mpam_resctrl_counters[eventid];
>> @@ -575,43 +558,38 @@ int resctrl_arch_cntr_read(struct rdt_resource *r, struct rdt_l3_mon_domain *d,
>>  	l3_dom = container_of(d, struct mpam_resctrl_dom, resctrl_mon_dom);
>>  	mon_comp = l3_dom->mon_comp[eventid];
>>  
>> -	return read_mon_cdp_safe(mon, mon_comp, mpam_feat_msmon_mbwu,
>> -				 USE_PRE_ALLOCATED, closid, rmid, val);
>> -}
>> -
>> -static void __reset_mon(struct mpam_resctrl_mon *mon, struct mpam_component *mon_comp,
>> -			int mon_idx,
>> -			enum resctrl_conf_type cdp_type, u32 closid, u32 rmid)
>> -{
>> -	struct mon_cfg cfg = { };
>> -
>> -	if (!mpam_is_enabled())
>> -		return;
>> +	if (cdp_enabled) {
>> +		u64 code_val = 0, data_val = 0;
>> +		int err;
>>  
>> -	/* Shift closid to account for CDP */
>> -	closid = resctrl_get_config_index(closid, cdp_type);
>> +		err = __read_mon(mon_comp, mpam_feat_msmon_mbwu,
>> +				 mon_for_cntr(mon, cntr_id, CDP_CODE),
>> +				 CDP_CODE, closid, rmid, &code_val);
>> +		if (err)
>> +			return err;
>>  
>> -	if (mon_idx == USE_PRE_ALLOCATED) {
>> -		int mbwu_idx = resctrl_arch_rmid_idx_encode(closid, rmid);
>> +		err = __read_mon(mon_comp, mpam_feat_msmon_mbwu,
>> +				 mon_for_cntr(mon, cntr_id, CDP_DATA),
>> +				 CDP_DATA, closid, rmid, &data_val);
>> +		if (err)
>> +			return err;
>>  
>> -		mon_idx = mon->mbwu_idx_to_mon[mbwu_idx];
>> +		*val = code_val + data_val;
>> +		return 0;
>>  	}
>>  
>> -	if (mon_idx == -1)
>> -		return;
>> -	cfg.mon = mon_idx;
>> -	mpam_msmon_reset_mbwu(mon_comp, &cfg);
>> +	return __read_mon(mon_comp, mpam_feat_msmon_mbwu,
>> +			  mon_for_cntr(mon, cntr_id, CDP_NONE),
>> +			  CDP_NONE, closid, rmid, val);
>>  }
>>  
>> -static void reset_mon_cdp_safe(struct mpam_resctrl_mon *mon, struct mpam_component *mon_comp,
>> -			       int mon_idx, u32 closid, u32 rmid)
>> +/* Reset the monitor resctrl's counter @cntr_id uses for @cdp_type. */
>> +static void __reset_mon(struct mpam_resctrl_mon *mon, struct mpam_component *mon_comp,
>> +			int cntr_id, enum resctrl_conf_type cdp_type)
>>  {
>> -	if (cdp_enabled) {
>> -		__reset_mon(mon, mon_comp, mon_idx, CDP_CODE, closid, rmid);
>> -		__reset_mon(mon, mon_comp, mon_idx, CDP_DATA, closid, rmid);
>> -	} else {
>> -		__reset_mon(mon, mon_comp, mon_idx, CDP_NONE, closid, rmid);
>> -	}
>> +	struct mon_cfg cfg = { .mon = mon_for_cntr(mon, cntr_id, cdp_type) };
>> +
>> +	mpam_msmon_reset_mbwu(mon_comp, &cfg);
>>  }
>>  
>>  /* Reset an assigned counter */
>> @@ -632,7 +610,12 @@ void resctrl_arch_reset_cntr(struct rdt_resource *r, struct rdt_l3_mon_domain *d
>>  	l3_dom = container_of(d, struct mpam_resctrl_dom, resctrl_mon_dom);
>>  	mon_comp = l3_dom->mon_comp[eventid];
>>  
>> -	reset_mon_cdp_safe(mon, mon_comp, USE_PRE_ALLOCATED, closid, rmid);
>> +	if (cdp_enabled) {
>> +		__reset_mon(mon, mon_comp, cntr_id, CDP_CODE);
>> +		__reset_mon(mon, mon_comp, cntr_id, CDP_DATA);
>> +	} else {
>> +		__reset_mon(mon, mon_comp, cntr_id, CDP_NONE);
>> +	}
>>  }
>>  
>>  /*
>> @@ -1170,22 +1153,6 @@ static void mpam_resctrl_pick_counters(void)
>>  	}
>>  }
>>  
>> -static void __config_cntr(struct mpam_resctrl_mon *mon, u32 cntr_id,
>> -			  enum resctrl_conf_type cdp_type, u32 closid, u32 rmid,
>> -			  bool assign)
>> -{
>> -	/* Same CDP index remap as closid; maps cntr_id to assigned_counters[]. */
>> -	u32 mbwu_idx, mon_idx = resctrl_get_config_index(cntr_id, cdp_type);
>> -
>> -	closid = resctrl_get_config_index(closid, cdp_type);
>> -	mbwu_idx = resctrl_arch_rmid_idx_encode(closid, rmid);
>> -
>> -	if (assign)
>> -		mon->mbwu_idx_to_mon[mbwu_idx] = mon->assigned_counters[mon_idx];
>> -	else
>> -		mon->mbwu_idx_to_mon[mbwu_idx] = -1;
>> -}
>> -
>>  void resctrl_arch_config_cntr(struct rdt_resource *r, struct rdt_l3_mon_domain *d,
>>  			      enum resctrl_event_id evtid, u32 rmid, u32 closid,
>>  			      u32 cntr_id, bool assign)
>> @@ -1197,19 +1164,13 @@ void resctrl_arch_config_cntr(struct rdt_resource *r, struct rdt_l3_mon_domain *
>>  		return;
>>  	}
>>  
>> -	if (!mon->mbwu_idx_to_mon || !mon->assigned_counters) {
>> -		pr_debug("monitor arrays not allocated\n");
>> +	if (!mon->assigned_counters) {
>> +		pr_debug("monitor array not allocated\n");
>>  		return;
>>  	}
>>  
>> -	if (cdp_enabled) {
>> -		__config_cntr(mon, cntr_id, CDP_CODE, closid, rmid, assign);
>> -		__config_cntr(mon, cntr_id, CDP_DATA, closid, rmid, assign);
>> -	} else {
>> -		__config_cntr(mon, cntr_id, CDP_NONE, closid, rmid, assign);
>> -	}
>> -
>> -	resctrl_arch_reset_cntr(r, d, closid, rmid, cntr_id, QOS_L3_MBM_TOTAL_EVENT_ID);
>> +	if (assign)
>> +		resctrl_arch_reset_cntr(r, d, closid, rmid, cntr_id, evtid);
>>  }
>>  
>>  static int mpam_resctrl_control_init(struct mpam_resctrl_res *res)
>> @@ -1296,25 +1257,16 @@ static int mpam_resctrl_pick_domain_id(int cpu, struct mpam_component *comp)
>>  static int mpam_resctrl_monitor_init_abmc(struct mpam_resctrl_mon *mon)
>>  {
>>  	struct mpam_resctrl_res *res = &mpam_resctrl_controls[RDT_RESOURCE_L3];
>> -	size_t num_rmid = resctrl_arch_system_num_rmid_idx();
>>  	struct rdt_resource *l3 = &res->resctrl_res;
>>  	struct mpam_class *class = mon->class;
>>  	u16 num_mbwu_mon;
>>  	int *cntrs;
>>  
>> -	int *rmid_array __free(kvfree) = kvmalloc_objs(*rmid_array, num_rmid);
>> -	if (!rmid_array) {
>> -		pr_debug("Failed to allocate RMID array\n");
>> -		return -ENOMEM;
>> -	}
>> -	memset(rmid_array, -1, num_rmid * sizeof(*rmid_array));
>> -
>>  	num_mbwu_mon = class->props.num_mbwu_mon;
>>  	cntrs = __alloc_mbwu_array(mon->class, num_mbwu_mon);
>>  	if (IS_ERR(cntrs))
>>  		return PTR_ERR(cntrs);
>>  	mon->assigned_counters = cntrs;
>> -	mon->mbwu_idx_to_mon = no_free_ptr(rmid_array);
>>  
>>  	l3->mon.mbm_cntr_assignable = true;
>>  	l3->mon.mbm_assign_on_mkdir = true;
>> @@ -1943,17 +1895,12 @@ static void mpam_resctrl_teardown_mon(struct mpam_resctrl_mon *mon, struct mpam_
>>  {
>>  	u32 num_mbwu_mon = class->props.num_mbwu_mon;
>>  
>> -	if (!mon->mbwu_idx_to_mon)
>> +	if (!mon->assigned_counters)
>>  		return;
>>  
>> -	if (mon->assigned_counters) {
>> -		__free_mbwu_mon(class, mon->assigned_counters, num_mbwu_mon);
>> -		kvfree(mon->assigned_counters);
>> -		mon->assigned_counters = NULL;
>> -	}
>> -
>> -	kvfree(mon->mbwu_idx_to_mon);
>> -	mon->mbwu_idx_to_mon = NULL;
>> +	__free_mbwu_mon(class, mon->assigned_counters, num_mbwu_mon);
>> +	kvfree(mon->assigned_counters);
>> +	mon->assigned_counters = NULL;
>>  }
>>  
>>  void mpam_resctrl_teardown_class(struct mpam_class *class)
> 



  reply	other threads:[~2026-11-17 11:48 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-13  8:35 [PATCH v2 0/2] fs/resctrl: Fix two counter ID bugs tizi4590
2026-11-13  8:35 ` [PATCH v2 1/2] arm_mpam: resctrl: Use the domain's counter ID tizi4590
2026-11-17 11:27   ` Ben Horgan
2026-11-17 11:48     ` Jinjie Ruan [this message]
2026-11-13  8:35 ` [PATCH v2 2/2] fs/resctrl: Give a monitor domain the counters the groups already use tizi4590
2026-11-13 16:14   ` Ben Horgan
2026-11-16  4:02     ` Jinjie Ruan
2026-11-17  3:27     ` Jinjie Ruan

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=f5cc83eb-7c67-4db0-add4-08e893640f92@huawei.com \
    --to=ruanjinjie@huawei.com \
    --cc=Dave.Martin@arm.com \
    --cc=babu.moger@amd.com \
    --cc=ben.horgan@arm.com \
    --cc=bp@alien8.de \
    --cc=fenghuay@nvidia.com \
    --cc=gshan@redhat.com \
    --cc=james.morse@arm.com \
    --cc=jic23@kernel.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=reinette.chatre@intel.com \
    --cc=tony.luck@intel.com \
    --cc=will@kernel.org \
    --cc=x86@kernel.org \
    /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