mbox series

[V4,00/14] genirq/msi: Spring cleaning

Message ID 20250319104921.201434198@linutronix.de
Headers show
Series genirq/msi: Spring cleaning | expand

Message

Thomas Gleixner March 19, 2025, 10:56 a.m. UTC
This is version 4 of the cleanup work. The previous version can be found
here:

   https://lore.kernel.org/all/20250317092919.008573387@linutronix.de

While converting the MSI descriptor locking to a lock guard() I stumbled
over various abuse of MSI descriptors (again).

The following series cleans up the offending code and converts the MSI
descriptor locking over to lock guards.

Changes vs. V3:

   - Cast retain_ptr() to void so it can't be used instead of return_ptr()
     - James

   - Split up the PCI/MSI changes

   - Move setting of pci_dev::msi_enabled to the success path

   - Fix a logic inversion in the UFS change and use a cleanup function to
     simplify the error path - James

   - Collect Reviewed/Tested/Acked-by tags where appropriate

The series applies on:

    git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git irq/msi

and is available from git:

    git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git irq/msi

Thanks,

	tglx
---
 drivers/ntb/msi.c                   |   22 +---
 drivers/pci/controller/pci-hyperv.c |   14 --
 drivers/pci/msi/api.c               |    6 -
 drivers/pci/msi/msi.c               |  175 ++++++++++++++++++++++--------------
 drivers/pci/pci.h                   |    9 +
 drivers/pci/tph.c                   |   44 ---------
 drivers/soc/ti/ti_sci_inta_msi.c    |   10 --
 drivers/ufs/host/ufs-qcom.c         |   85 +++++++++--------
 include/linux/cleanup.h             |   16 +++
 include/linux/irqdomain.h           |    2 
 include/linux/msi.h                 |    7 +
 kernel/irq/msi.c                    |  125 ++++++++++---------------
 12 files changed, 258 insertions(+), 257 deletions(-)

Comments

James Bottomley March 19, 2025, 8:40 p.m. UTC | #1
On Wed, 2025-03-19 at 11:57 +0100, Thomas Gleixner wrote:
> The driver abuses the MSI descriptors for internal purposes. Aside of
> core code and MSI providers nothing has to care about their
> existence. They have been encapsulated with a lot of effort because
> this kind of abuse caused all sorts of issues including a
> maintainability nightmare.
> 
> Rewrite the code so it uses dedicated storage to hand the required
> information to the interrupt handler and use a custom cleanup
> function to simplify the error path.
> 
> No functional change intended.
> 
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> Cc: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
> Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
> Cc: linux-scsi@vger.kernel.org
> ---
> V4: Fix inverse NULL pointer check and use __free() for error paths -
> James
> ---
>  drivers/ufs/host/ufs-qcom.c |   85 ++++++++++++++++++++++++---------
> -----------
>  1 file changed, 48 insertions(+), 37 deletions(-)
> 
> --- a/drivers/ufs/host/ufs-qcom.c
> +++ b/drivers/ufs/host/ufs-qcom.c
> @@ -1782,25 +1782,38 @@ static void ufs_qcom_write_msi_msg(struc
>  	ufshcd_mcq_config_esi(hba, msg);
>  }
>  
> +struct ufs_qcom_irq {
> +	unsigned int		irq;
> +	unsigned int		idx;
> +	struct ufs_hba		*hba;
> +};
> +
>  static irqreturn_t ufs_qcom_mcq_esi_handler(int irq, void *data)
>  {
> -	struct msi_desc *desc = data;
> -	struct device *dev = msi_desc_to_dev(desc);
> -	struct ufs_hba *hba = dev_get_drvdata(dev);
> -	u32 id = desc->msi_index;
> -	struct ufs_hw_queue *hwq = &hba->uhq[id];
> +	struct ufs_qcom_irq *qi = data;
> +	struct ufs_hba *hba = qi->hba;
> +	struct ufs_hw_queue *hwq = &hba->uhq[qi->idx];
>  
> -	ufshcd_mcq_write_cqis(hba, 0x1, id);
> +	ufshcd_mcq_write_cqis(hba, 0x1, qi->idx);
>  	ufshcd_mcq_poll_cqe_lock(hba, hwq);
>  
>  	return IRQ_HANDLED;
>  }
>  
> +static void ufs_qcom_irq_free(struct ufs_qcom_irq *uqi)
> +{
> +	for (struct ufs_qcom_irq *q = uqi; q->irq; q++)
> +		devm_free_irq(q->hba->dev, q->irq, q->hba);
> +
> +	platform_device_msi_free_irqs_all(uqi->hba->dev);
> +	devm_kfree(uqi->hba->dev, uqi);
> +}
> +
> +DEFINE_FREE(ufs_qcom_irq, struct ufs_qcom_irq *, if (_T)
> ufs_qcom_irq_free(_T))

So if every __free body basically has a condition check and an external
call, it would be a lot nicer if the macro were coded as something like

#define DEFINE_FREE(_name, _type, _cond, _free) \
	static inline void __free_##_name(void *p) { _type _T =
*(_type *)p; if (_cond(_T)) _free(_T); }

to avoid having to splash the strange _T across the kernel.

Other than that (which isn't your problem, so you can ignore it), the
code looks fine but I can't test it.

Reviewed-by: James Bottomley <James.Bottomley@HansenPartnership.com>