From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by smtp.subspace.kernel.org (Postfix) with ESMTP id 9B0A6345ED5; Thu, 23 Jul 2026 10:20:29 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=217.140.110.172 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784802031; cv=none; b=qws4d2pGWcMVBUJ/Db75n76pAntgVDD7jg+LHW9kAEiMnptcUm6hKVQ6hL3idFPbQLQ0paBCrMCo263tstHz362c0ZY6g8OyTJ95xabteUnUqeWaaSiYpJ83ldCQv1F6lRJw3qUFlfZxPh8zgBULB68y9VF+hS455Hu6+BOiEgc= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784802031; c=relaxed/simple; bh=DEwZHcaCuIqEH85LWq1xXjbGMYiVUhbTerLFXP69HkQ=; h=Message-ID:Date:MIME-Version:Subject:To:Cc:References:From: In-Reply-To:Content-Type; b=Arxy9/4pdFwJDyzb9LykFeOUngASe+Z4P+555Eyv5KPlV2RZ227RjrJziMHc71KSjduY6jrXIXjrV0l/8lqxC/Jl1EbCS9rSorzJG3XUPgOPiZZQO/EB/OUTMO12MTPd6lhzVh+wTIbG8wyNAISeK698asYNawOamVSYQ7mcNuI= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=arm.com; spf=pass smtp.mailfrom=arm.com; dkim=pass (1024-bit key) header.d=arm.com header.i=@arm.com header.b=KEDQDiaE; arc=none smtp.client-ip=217.140.110.172 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=arm.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=arm.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=arm.com header.i=@arm.com header.b="KEDQDiaE" Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 7CC361477; Thu, 23 Jul 2026 03:20:24 -0700 (PDT) Received: from [10.2.212.8] (e134344.arm.com [10.2.212.8]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 0419B3F86F; Thu, 23 Jul 2026 03:20:25 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=arm.com; s=foss; t=1784802028; bh=DEwZHcaCuIqEH85LWq1xXjbGMYiVUhbTerLFXP69HkQ=; h=Date:Subject:To:Cc:References:From:In-Reply-To:From; b=KEDQDiaEKPA/TdWauC8U9sHV45XGgwSTYi9Ko6GB+pKc5mcgbBhBa45j3CkH9ouOK WLB4mSl+c98YzV4FJ2+2In7xkEVwZYUxGQlL/+6A17rIB284TpDJ/2P0/LKqrLYZ5P tXjil7qNGxzQQFWbwxaXIDq/XQ7WGrkOGbtUR3BY= Message-ID: Date: Thu, 23 Jul 2026 11:20:24 +0100 Precedence: bulk X-Mailing-List: linux-acpi@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 User-Agent: Thunderbird Daily Subject: Re: [PATCH v3 07/16] arm_mpam: __ris_msmon_read(): get rid of nrdy special handling To: Andre Przywara , Lorenzo Pieralisi , Hanjun Guo , Sudeep Holla , Catalin Marinas , Will Deacon , "Rafael J . Wysocki" , Len Brown , James Morse , Reinette Chatre , Fenghua Yu Cc: Jonathan Cameron , Srivathsa L Rao , Ganapatrao Kulkarni , Trilok Soni , Srinivas Ramana , Niyas Sait , linux-acpi@vger.kernel.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org References: <20260710144520.917375-1-andre.przywara@arm.com> <20260710144520.917375-8-andre.przywara@arm.com> <778a9fc1-cc11-4d7f-b5a5-78a1740520a1@arm.com> Content-Language: en-US From: Ben Horgan In-Reply-To: Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Hi Andre, On 7/23/26 10:45, Andre Przywara wrote: > Hi Ben, > > On 7/20/26 18:09, Ben Horgan wrote: >> Hi Andre, >> >> On 7/20/26 16:58, Andre Przywara wrote: >>> Hi Ben, >>> >>> thanks for having a look! >>> >>> On 7/15/26 15:39, Ben Horgan wrote: >>>> Hi Andre, >>>> >>>> On 7/10/26 15:45, Andre Przywara wrote: >>>>> Although so far MSC accesses couldn't fail, there is one special >>>>> condition that would create an error: when the MBWU counter wouldn't be >>>>> able to read a stable value, we were setting bit 63 to mark this value >>>>> as unstable, and return this as an error later. >>>>> Now since the functions can return a proper error value, we can get rid of >>>>> this kludge and use the return value directly. >>>>> >>>>> Remove the "nrdy" error flag variable, and assign -EBUSY to "ret" to handle >>>>> this case. >>>> >>>> I don't think we want this patch. The h/w can still return (as much as it ever could) and so we >>>> still need to handle it even if we are no longer augmenting its meaning in software to also >>>> indicate >>>> an unstable 64 bit value. >>> >>> Mmh, not sure I understand your concern: to me it looks like nrdy is some kind of error flag, that >>> we used in absence of a proper error return value. Now we have "int ret;", so can use that directly? >>> But to me it looks like nothing really changes, or did I miss something? >>> >>> I have no really strong opinion of this patch, it was more an pportunity to consolidate the crude >>> error handling in this function. I am happy to drop it, if you like, maybe we can revisit this >>> later. >> >> What I was trying to say is that mpam_msc_read_mbwu_l() could previously return a value with bit 63, >> MSMON__L_NRDY set in two cases, one set by s/w and one set by h/w. Either when it reads that >> directly from the hardware or when it is set in the function to indicate an unstable value. The h/w >> case is the same for 31 bit counters too except in that case the h/w sets bit 31, MSMON_NRDY. Using >> 'ret' to directly return -EBUSY for the s/w case where a stable value is not reached for 44 or 63 >> bit counters doesn't mean that the h/w case won't happen. > > I am not sure I see the problem, the idea of this patch was to use the opportunity of having now a > proper return value, and to not hide that "nrdy" is actually an error flag. If I read the code > correctly, then at the moment we flag the error early (using nrdy), but then continue with > (potentially bogus?) "now" calculations, only to discard them towards the end of the function, to > return an error when nrdy was set. So my idea was to just handle the error case early and return. > Or do you mean I was just missing one case where NRDY was set? I think the problem comes in patch 4 actually. Where you remove the checking for L_NRDY, which can still be read from h/w. Both long and 31 bit counters would need to be considered for this kind of cleanup. > > In any case, to not jeopardise the whole series over this rather opportunistic patch, I will just > drop any changes to nrdy handling. This makes the remaining patches easier to understand, I guess, > since they are now more or less schematic "if (err) return err;" changes. > > I think we can clean this up later if needed, in a follow up patch. Sure. Thanks, Ben > > Thanks, > Andre > >>>>> >>>>> Signed-off-by: Andre Przywara >>>>> --- >>>>>    drivers/resctrl/mpam_devices.c | 38 +++++++++++++++------------------- >>>>>    1 file changed, 17 insertions(+), 21 deletions(-) >>>>> >>>>> diff --git a/drivers/resctrl/mpam_devices.c b/drivers/resctrl/mpam_devices.c >>>>> index 84a8715464be..530ac0fe97b5 100644 >>>>> --- a/drivers/resctrl/mpam_devices.c >>>>> +++ b/drivers/resctrl/mpam_devices.c >>>>> @@ -1306,7 +1306,6 @@ static void __ris_msmon_read(void *arg) >>>>>        u64 now; >>>>>        int ret; >>>>>        u32 now32; >>>>> -    bool nrdy = false; >>>>>        bool config_mismatch; >>>>>        bool overflow = false; >>>>>        struct mon_read *m = arg; >>>>> @@ -1371,14 +1370,18 @@ static void __ris_msmon_read(void *arg) >>>>>        switch (m->type) { >>>>>        case mpam_feat_msmon_csu: >>>>>            ret = mpam_read_monsel_reg(msc, CSU, &now32); >>>>> +        if (!ret) { >>>>> +            if ((now32 & MSMON___NRDY)) >>>>> +                ret = -EBUSY; >>>>> + >>>>> +            if (mpam_has_quirk(IGNORE_CSU_NRDY, msc) && >>>>> +                m->waited_timeout) >>>>> +                ret = 0; >>>>> +        } >>>>>            if (ret) >>>>>                goto out_unlock; >>>>> -        nrdy = now32 & MSMON___NRDY; >>>>> -        now = FIELD_GET(MSMON___VALUE, now32); >>>>> - >>>>> -        if (mpam_has_quirk(IGNORE_CSU_NRDY, msc) && m->waited_timeout) >>>>> -            nrdy = false; >>>>>    +        now = FIELD_GET(MSMON___VALUE, now32); >>>>>            break; >>>>>        case mpam_feat_msmon_mbwu_31counter: >>>>>        case mpam_feat_msmon_mbwu_44counter: >>>>> @@ -1394,9 +1397,11 @@ static void __ris_msmon_read(void *arg) >>>>>                    now = FIELD_GET(MSMON___L_VALUE, now); >>>>>            } else { >>>>>                ret = mpam_read_monsel_reg(msc, MBWU, &now32); >>>>> +            if (!ret && (now32 & MSMON___NRDY)) >>>>> +                ret = -EBUSY; >>>>>                if (ret) >>>>>                    goto out_unlock; >>>>> -            nrdy = now32 & MSMON___NRDY; >>>>> + >>>>>                now = FIELD_GET(MSMON___VALUE, now32); >>>>>            } >>>>>    @@ -1404,9 +1409,6 @@ static void __ris_msmon_read(void *arg) >>>>>                m->type != mpam_feat_msmon_mbwu_63counter) >>>>>                now *= 64; >>>>>    -        if (nrdy) >>>>> -            break; >>>>> - >>>>>            mbwu_state = &ris->mbwu_state[ctx->mon]; >>>>>              if (overflow) >>>>> @@ -1419,22 +1421,16 @@ static void __ris_msmon_read(void *arg) >>>>>            now += mbwu_state->correction; >>>>>            break; >>>>>        default: >>>>> -        m->err = -EINVAL; >>>>> +        ret = -EINVAL; >>>>>        } >>>>> -    mpam_mon_sel_unlock(msc); >>>>> - >>>>> -    if (nrdy) >>>>> -        m->err = -EBUSY; >>>>> - >>>>> -    if (!m->err) >>>>> -        *m->val += now; >>>>> - >>>>> -    return; >>>>>      out_unlock: >>>>>        mpam_mon_sel_unlock(msc); >>>>>    -    m->err = ret; >>>>> +    if (ret) >>>>> +        m->err = ret; >>>>> +    else >>>>> +        *m->val += now; >>>>>    } >>>>>      static int _msmon_read(struct mpam_component *comp, struct mon_read *arg) >>>> >>> >> >