Re: [PATCH net-next v8 3/3] net: stmmac: Add driver support for DWMAC5 common safety IRQ

From: Suraj Jaiswal
Date: Wed Dec 27 2023 - 06:04:30 EST

Next message: Barry Song: "Re: [PATCH v2] crypto: scompress - fix req->dst buffer overflow"
Previous message: JeeHeng Sia: "RE: [RFC 00/16] Basic clock and reset support for StarFive JH8100 RISC-V SoC"
In reply to: Suraj Jaiswal: "Re: [PATCH net-next v8 3/3] net: stmmac: Add driver support for DWMAC5 common safety IRQ"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Seren,
please find the updated comment .

Thanks
Suraj

On 12/26/2023 4:40 PM, Suraj Jaiswal wrote:
> Hi seren
> let me check below on test setup once & get back
>
> Thanks
> Suraj
>
> On 12/22/2023 8:05 PM, Serge Semin wrote:
>> On Fri, Dec 22, 2023 at 02:13:49PM +0530, Suraj Jaiswal wrote:
>>> HI Serge,
>>> please find commnet inline.
>>>
>>> Thanks
>>> Suraj
>>>
>>> On 12/21/2023 6:19 PM, Serge Semin wrote:
>>>> Hi Suraj
>>>>
>>>> On Thu, Dec 21, 2023 at 01:06:20PM +0530, Suraj Jaiswal wrote:
>>>>> Add support to listen HW safety IRQ like ECC(error
>>>>> correction code), DPP(data path parity), FSM(finite state
>>>>> machine) fault in common IRQ line.
>>>>>
>>>>> Signed-off-by: Suraj Jaiswal <quic_jsuraj@xxxxxxxxxxx>
>>>>
>>>> Thanks for taking my notes into account. One more comment is further
>>>> below.
>>>>
>>>>> ---
>>>>> drivers/net/ethernet/stmicro/stmmac/common.h | 1 +
>>>>> drivers/net/ethernet/stmicro/stmmac/stmmac.h | 3 ++
>>>>> .../net/ethernet/stmicro/stmmac/stmmac_main.c | 37 +++++++++++++++++++
>>>>> .../ethernet/stmicro/stmmac/stmmac_platform.c | 8 ++++
>>>>> 4 files changed, 49 insertions(+)
>>>>>
>>>>> diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
>>>>> index 721c1f8e892f..b9233b09b80f 100644
>>>>> --- a/drivers/net/ethernet/stmicro/stmmac/common.h
>>>>> +++ b/drivers/net/ethernet/stmicro/stmmac/common.h
>>>>> @@ -344,6 +344,7 @@ enum request_irq_err {
>>>>> REQ_IRQ_ERR_ALL,
>>>>> REQ_IRQ_ERR_TX,
>>>>> REQ_IRQ_ERR_RX,
>>>>> + REQ_IRQ_ERR_SFTY,
>>>>> REQ_IRQ_ERR_SFTY_UE,
>>>>> REQ_IRQ_ERR_SFTY_CE,
>>>>> REQ_IRQ_ERR_LPI,
>>>>> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
>>>>> index 9f89acf31050..ca3d93851bed 100644
>>>>> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
>>>>> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
>>>>> @@ -31,6 +31,7 @@ struct stmmac_resources {
>>>>> int wol_irq;
>>>>> int lpi_irq;
>>>>> int irq;
>>>>> + int sfty_irq;
>>>>> int sfty_ce_irq;
>>>>> int sfty_ue_irq;
>>>>> int rx_irq[MTL_MAX_RX_QUEUES];
>>>>> @@ -297,6 +298,7 @@ struct stmmac_priv {
>>>>> void __iomem *ptpaddr;
>>>>> void __iomem *estaddr;
>>>>> unsigned long active_vlans[BITS_TO_LONGS(VLAN_N_VID)];
>>>>> + int sfty_irq;
>>>>> int sfty_ce_irq;
>>>>> int sfty_ue_irq;
>>>>> int rx_irq[MTL_MAX_RX_QUEUES];
>>>>> @@ -305,6 +307,7 @@ struct stmmac_priv {
>>>>> char int_name_mac[IFNAMSIZ + 9];
>>>>> char int_name_wol[IFNAMSIZ + 9];
>>>>> char int_name_lpi[IFNAMSIZ + 9];
>>>>> + char int_name_sfty[IFNAMSIZ + 10];
>>>>> char int_name_sfty_ce[IFNAMSIZ + 10];
>>>>> char int_name_sfty_ue[IFNAMSIZ + 10];
>>>>> char int_name_rx_irq[MTL_MAX_TX_QUEUES][IFNAMSIZ + 14];
>>>>> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>>>>> index 47de466e432c..7d4e827dfeab 100644
>>>>> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>>>>> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>>>>> @@ -3592,6 +3592,10 @@ static void stmmac_free_irq(struct net_device *dev,
>>>>> if (priv->wol_irq > 0 && priv->wol_irq != dev->irq)
>>>>> free_irq(priv->wol_irq, dev);
>>>>> fallthrough;
>>>>> + case REQ_IRQ_ERR_SFTY:
>>>>> + if (priv->sfty_irq > 0 && priv->sfty_irq != dev->irq)
>>>>> + free_irq(priv->sfty_irq, dev);
>>>>> + fallthrough;
>>>>> case REQ_IRQ_ERR_WOL:
>>>>> free_irq(dev->irq, dev);
>>>>> fallthrough;
>>>>> @@ -3661,6 +3665,23 @@ static int stmmac_request_irq_multi_msi(struct net_device *dev)
>>>>> }
>>>>> }
>>>>>
>>>>> + /* Request the common Safety Feature Correctible/Uncorrectible
>>>>> + * Error line in case of another line is used
>>>>> + */
>>>>> + if (priv->sfty_irq > 0 && priv->sfty_irq != dev->irq) {
>>>>> + int_name = priv->int_name_sfty;
>>>>> + sprintf(int_name, "%s:%s", dev->name, "safety");
>>>>> + ret = request_irq(priv->sfty_irq, stmmac_safety_interrupt,
>>>>> + 0, int_name, dev);
>>>>> + if (unlikely(ret < 0)) {
>>>>> + netdev_err(priv->dev,
>>>>> + "%s: alloc sfty MSI %d (error: %d)\n",
>>>>> + __func__, priv->sfty_irq, ret);
>>>>> + irq_err = REQ_IRQ_ERR_SFTY;
>>>>> + goto irq_error;
>>>>> + }
>>>>> + }
>>>>> +
>>>>> /* Request the Safety Feature Correctible Error line in
>>>>> * case of another line is used
>>>>> */
>>>>> @@ -3798,6 +3819,21 @@ static int stmmac_request_irq_single(struct net_device *dev)
>>>>> }
>>>>> }
>>>>>
>>>>> + /* Request the common Safety Feature Correctible/Uncorrectible
>>>>> + * Error line in case of another line is used
>>>>> + */
>>>>> + if (priv->sfty_irq > 0 && priv->sfty_irq != dev->irq) {
>>>>
>>>>> + ret = request_irq(priv->sfty_irq, stmmac_safety_interrupt,
>>>>> + IRQF_SHARED, dev->name, dev);
>>>>
>>>> Just noticed yesterday that stmmac_safety_interrupt() is also called
>>>> from the stmmac_interrupt() handler which is supposed to be registered
>>>> on the generic "mac" IRQ. Won't it cause races around the CSRs
>>>> (doubtfully but still worth to note) and the errors handling
>>>> (stmmac_global_err()) in case if both IRQs are raised simultaneously?
>>>> At the very least it looks suspicious and worth double-checking.
>>>>
>>>> I also found out that nobody seemed to care that the same handler is
>>>> registered on MAC, WoL and LPI IRQ lines. Hmm, no race-related
>>>> problems have been reported so far for the platforms with separate
>>>> WoL/LPI IRQs. It's either a lucky coincident or the IRQs are always
>>>> assigned to the same CPU or the IRQs handle is indeed free of races.
>>>> In anyway it looks suspicious too. At the very least AFAICS the DMA
>>>> IRQ-handler is indeed racy on the status CSR access. It isn't
>>>> cleared-on-read, but write-one-to-clear. So the statistics might be
>>>> calculated more than once for the same CSR state. There might be some
>>>> other problems I failed to spot on the first glance.
>>>>
>>>> David, Eric, Jacub, Paolo, your opinion about the note above?
>>>>
>>>> -Serge(y)
>>>>
>>
>>> <Suraj> We are adding common IRQ similar to already present code for correcteable/uncorrecable https://elixir.bootlin.com/linux/latest/source/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c#L3592.
>>
>> From that perspective your change in stmmac_request_irq_multi_msi() is
>> correct, but stmmac_request_irq_single() is another story. The first
>> one method implies assigning the individual IRQ handlers to all
>> available lines. The later method assigns the _common_ handler to all
>> the lines. The common handler already calls the Safety IRQ handler -
>> stmmac_safety_feat_interrupt(). So should the safety IRQ line is
>> separately available it's possible to have the Safety IRQ handlers
>> executed concurrently - in framework of the common IRQ events handling
>> (if safety IRQ is raised during the common IRQ being handled) and
>> individual Safety IRQ. It's prune to the race condition I pointed out
>> to in my message above. Did you consider that problem?
>>
>>> Also, we need the sfty IRQ handling as soon as the fault occured & that can only be handled if we have handler attached with sfty IRQ.
>>> stmmac_interrupt() will only be triggerd when interrupt triggered for rx/tx packet .
>>> while registerting with sfty IRQ will get triggered as soon as emac HW detect the fault.
>>
>> Please read my comment more carefully. The safety IRQ can be raised
>> during the common IRQ handling, thus the
>> stmmac_safety_feat_interrupt() method might get to be concurrently
>> executed.
>>
>> -Serge(y)
>>
<Suraj> Have testing this on device . We have added print in the both the places stmmac_interrupt() as well as sfty interrupt handler.
We can see that sfty interrupt handler is getting triggred first & stmmac_safety_feat_interrupt () code added in stmmac_intterupt() is not getting triggred because looks like interrupt status bit register is already getting cleared as part of sfty interrupt handler. So it looks good . Please let us know if any further comment.

Please find the log below .

/ # [ 1505.602173] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1505.607274] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'RXCES: MTL RX Memory Error'
[ 1505.617395] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1505.622494] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'TXCES: MTL TX Memory Error'
[ 1505.888913] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1505.894010] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'RXCES: MTL RX Memory Error'
[ 1506.605821] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1506.610919] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'RXCES: MTL RX Memory Error'
[ 1506.621034] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1506.626131] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'TXCES: MTL TX Memory Error'
[ 1507.613036] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1507.618133] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'RXCES: MTL RX Memory Error'
[ 1507.628249] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1507.633346] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'TXCES: MTL TX Memory Error'
[ 1508.619034] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1508.624132] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'RXCES: MTL RX Memory Error'
[ 1508.634245] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1508.639343] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'TXCES: MTL TX Memory Error'
[ 1509.631151] sj: stmmac_safety_interrupt from sfty IRQ handler
[ 1509.636249] qcom-ethqos 23040000.ethernet eth1: Found correctable error in MTL: 'RXCES: MTL RX Memory Error'

>>>
>>>>> + if (unlikely(ret < 0)) {
>>>>> + netdev_err(priv->dev,
>>>>> + "%s: ERROR: allocating the sfty IRQ %d (%d)\n",
>>>>> + __func__, priv->sfty_irq, ret);
>>>>> + irq_err = REQ_IRQ_ERR_SFTY;
>>>>> + goto irq_error;
>>>>> + }
>>>>> + }
>>>>> +
>>>>> return 0;
>>>>>
>>>>> irq_error:
>>>>> @@ -7462,6 +7498,7 @@ int stmmac_dvr_probe(struct device *device,
>>>>> priv->dev->irq = res->irq;
>>>>> priv->wol_irq = res->wol_irq;
>>>>> priv->lpi_irq = res->lpi_irq;
>>>>> + priv->sfty_irq = res->sfty_irq;
>>>>> priv->sfty_ce_irq = res->sfty_ce_irq;
>>>>> priv->sfty_ue_irq = res->sfty_ue_irq;
>>>>> for (i = 0; i < MTL_MAX_RX_QUEUES; i++)
>>>>> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
>>>>> index 70eadc83ca68..ab250161fd79 100644
>>>>> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
>>>>> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
>>>>> @@ -743,6 +743,14 @@ int stmmac_get_platform_resources(struct platform_device *pdev,
>>>>> dev_info(&pdev->dev, "IRQ eth_lpi not found\n");
>>>>> }
>>>>>
>>>>> + stmmac_res->sfty_irq =
>>>>> + platform_get_irq_byname_optional(pdev, "sfty");
>>>>> + if (stmmac_res->sfty_irq < 0) {
>>>>> + if (stmmac_res->sfty_irq == -EPROBE_DEFER)
>>>>> + return -EPROBE_DEFER;
>>>>> + dev_info(&pdev->dev, "IRQ safety IRQ not found\n");
>>>>> + }
>>>>> +
>>>>> stmmac_res->addr = devm_platform_ioremap_resource(pdev, 0);
>>>>>
>>>>> return PTR_ERR_OR_ZERO(stmmac_res->addr);
>>>>> --
>>>>> 2.25.1
>>>>>
>>>>>
>

Next message: Barry Song: "Re: [PATCH v2] crypto: scompress - fix req->dst buffer overflow"
Previous message: JeeHeng Sia: "RE: [RFC 00/16] Basic clock and reset support for StarFive JH8100 RISC-V SoC"
In reply to: Suraj Jaiswal: "Re: [PATCH net-next v8 3/3] net: stmmac: Add driver support for DWMAC5 common safety IRQ"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]