public inbox for linux-s390@vger.kernel.org
 help / color / mirror / Atom feed
From: Farhan Ali <alifm@linux.ibm.com>
To: "Cédric Le Goater" <clg@redhat.com>,
	linux-s390@vger.kernel.org, kvm@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org
Cc: alex.williamson@redhat.com, helgaas@kernel.org,
	schnelle@linux.ibm.com, mjrosato@linux.ibm.com
Subject: Re: [PATCH v3 03/10] PCI: Allow per function PCI slots
Date: Tue, 16 Sep 2025 11:37:54 -0700	[thread overview]
Message-ID: <e86caff6-8af0-48c9-9058-c1991e23160f@linux.ibm.com> (raw)
In-Reply-To: <07205677-09f0-464b-b31c-0fb5493a1d81@redhat.com>


On 9/15/2025 11:52 PM, Cédric Le Goater wrote:
> Hello Ali,
>
> On 9/11/25 20:33, Farhan Ali wrote:
>> On s390 systems, which use a machine level hypervisor, PCI devices are
>> always accessed through a form of PCI pass-through which fundamentally
>> operates on a per PCI function granularity. This is also reflected in 
>> the
>> s390 PCI hotplug driver which creates hotplug slots for individual PCI
>> functions. Its reset_slot() function, which is a wrapper for
>> zpci_hot_reset_device(), thus also resets individual functions.
>>
>> Currently, the kernel's PCI_SLOT() macro assigns the same pci_slot 
>> object
>> to multifunction devices. This approach worked fine on s390 systems that
>> only exposed virtual functions as individual PCI domains to the 
>> operating
>> system.  Since commit 44510d6fa0c0 ("s390/pci: Handling multifunctions")
>> s390 supports exposing the topology of multifunction PCI devices by
>> grouping them in a shared PCI domain. When attempting to reset a 
>> function
>> through the hotplug driver, the shared slot assignment causes the wrong
>> function to be reset instead of the intended one. It also leaks 
>> memory as
>> we do create a pci_slot object for the function, but don't correctly 
>> free
>> it in pci_slot_release().
>>
>> Add a flag for struct pci_slot to allow per function PCI slots for
>> functions managed through a hypervisor, which exposes individual PCI
>> functions while retaining the topology.
>>
>> Fixes: 44510d6fa0c0 ("s390/pci: Handling multifunctions")
>> Suggested-by: Niklas Schnelle <schnelle@linux.ibm.com>
>> Signed-off-by: Farhan Ali <alifm@linux.ibm.com>
>> ---
>>   drivers/pci/hotplug/s390_pci_hpc.c | 10 ++++++++--
>>   drivers/pci/pci.c                  |  4 +++-
>>   drivers/pci/slot.c                 | 14 +++++++++++---
>>   include/linux/pci.h                |  1 +
>>   4 files changed, 23 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/pci/hotplug/s390_pci_hpc.c 
>> b/drivers/pci/hotplug/s390_pci_hpc.c
>> index d9996516f49e..8b547de464bf 100644
>> --- a/drivers/pci/hotplug/s390_pci_hpc.c
>> +++ b/drivers/pci/hotplug/s390_pci_hpc.c
>> @@ -126,14 +126,20 @@ static const struct hotplug_slot_ops 
>> s390_hotplug_slot_ops = {
>>     int zpci_init_slot(struct zpci_dev *zdev)
>>   {
>> +    int ret;
>>       char name[SLOT_NAME_SIZE];
>>       struct zpci_bus *zbus = zdev->zbus;
>>         zdev->hotplug_slot.ops = &s390_hotplug_slot_ops;
>>         snprintf(name, SLOT_NAME_SIZE, "%08x", zdev->fid);
>> -    return pci_hp_register(&zdev->hotplug_slot, zbus->bus,
>> -                   zdev->devfn, name);
>> +    ret = pci_hp_register(&zdev->hotplug_slot, zbus->bus,
>> +                zdev->devfn, name);
>> +    if (ret)
>> +        return ret;
>> +
>> +    zdev->hotplug_slot.pci_slot->per_func_slot = 1;
>> +    return 0;
>>   }
>>     void zpci_exit_slot(struct zpci_dev *zdev)
>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>> index 3994fa82df68..70296d3b1cfc 100644
>> --- a/drivers/pci/pci.c
>> +++ b/drivers/pci/pci.c
>> @@ -5061,7 +5061,9 @@ static int pci_reset_hotplug_slot(struct 
>> hotplug_slot *hotplug, bool probe)
>>     static int pci_dev_reset_slot_function(struct pci_dev *dev, bool 
>> probe)
>>   {
>> -    if (dev->multifunction || dev->subordinate || !dev->slot ||
>> +    if (dev->multifunction && !dev->slot->per_func_slot)
>> +        return -ENOTTY;
>> +    if (dev->subordinate || !dev->slot ||
>>           dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
>>           return -ENOTTY;
>>   diff --git a/drivers/pci/slot.c b/drivers/pci/slot.c
>> index 50fb3eb595fe..51ee59e14393 100644
>> --- a/drivers/pci/slot.c
>> +++ b/drivers/pci/slot.c
>> @@ -63,6 +63,14 @@ static ssize_t cur_speed_read_file(struct pci_slot 
>> *slot, char *buf)
>>       return bus_speed_read(slot->bus->cur_bus_speed, buf);
>>   }
>>   +static bool pci_dev_matches_slot(struct pci_dev *dev, struct 
>> pci_slot *slot)
>> +{
>> +    if (slot->per_func_slot)
>> +        return dev->devfn == slot->number;
>> +
>> +    return PCI_SLOT(dev->devfn) == slot->number;
>> +}
>> +
>>   static void pci_slot_release(struct kobject *kobj)
>>   {
>>       struct pci_dev *dev;
>> @@ -73,7 +81,7 @@ static void pci_slot_release(struct kobject *kobj)
>>         down_read(&pci_bus_sem);
>>       list_for_each_entry(dev, &slot->bus->devices, bus_list)
>> -        if (PCI_SLOT(dev->devfn) == slot->number)
>> +        if (pci_dev_matches_slot(dev, slot))
>>               dev->slot = NULL;
>>       up_read(&pci_bus_sem);
>>   @@ -166,7 +174,7 @@ void pci_dev_assign_slot(struct pci_dev *dev)
>>         mutex_lock(&pci_slot_mutex);
>>       list_for_each_entry(slot, &dev->bus->slots, list)
>> -        if (PCI_SLOT(dev->devfn) == slot->number)
>> +        if (pci_dev_matches_slot(dev, slot))
>>               dev->slot = slot;
>>       mutex_unlock(&pci_slot_mutex);
>>   }
>> @@ -285,7 +293,7 @@ struct pci_slot *pci_create_slot(struct pci_bus 
>> *parent, int slot_nr,
>>         down_read(&pci_bus_sem);
>>       list_for_each_entry(dev, &parent->devices, bus_list)
>> -        if (PCI_SLOT(dev->devfn) == slot_nr)
>> +        if (pci_dev_matches_slot(dev, slot))
>>               dev->slot = slot;
>>       up_read(&pci_bus_sem);
>>   diff --git a/include/linux/pci.h b/include/linux/pci.h
>> index 59876de13860..9265f32d9786 100644
>> --- a/include/linux/pci.h
>> +++ b/include/linux/pci.h
>> @@ -78,6 +78,7 @@ struct pci_slot {
>>       struct list_head    list;        /* Node in list of slots */
>>       struct hotplug_slot    *hotplug;    /* Hotplug info (move here) */
>>       unsigned char        number;        /* PCI_SLOT(pci_dev->devfn) */
>> +    unsigned int        per_func_slot:1; /* Allow per function slot */
>>       struct kobject        kobj;
>>   };
>
> This change generates a kernel oops on x86_64. It can be reproduced in 
> a VM.
>
>
> C.
>
> [    3.073990] BUG: kernel NULL pointer dereference, address: 
> 0000000000000021
> [    3.074976] #PF: supervisor read access in kernel mode
> [    3.074976] #PF: error_code(0x0000) - not-present page
> [    3.074976] PGD 0 P4D 0
> [    3.074976] Oops: Oops: 0000 [#1] SMP NOPTI
> [    3.074976] CPU: 18 UID: 0 PID: 1 Comm: swapper/0 Not tainted 
> 6.17.0-rc6-clg-dirty #8 PREEMPT(voluntary)
> [    3.074976] Hardware name: Supermicro Super Server/X13SAE-F, BIOS 
> 4.2 12/17/2024
> [    3.074976] RIP: 0010:pci_reset_bus_function+0xdf/0x160
> [    3.074976] Code: 4e 08 00 00 40 0f 85 83 00 00 00 48 8b 78 18 e8 
> 27 9d ff ff 83 f8 e7 74 17 48 83 c4 08 5b 5d 41 5c c3 cc cc cc cc 48 
> 8b 43 30 <f6> 40 21 01 75 b6 48 8b 53 10 48 83 7a 10 00 74 5e 48 83 7b 
> 18 00
> [    3.074976] RSP: 0000:ffffcd808007b9a8 EFLAGS: 00010202
> [    3.074976] RAX: 0000000000000000 RBX: ffff88c4019b8000 RCX: 
> 0000000000000000
> [    3.074976] RDX: 0000000000000000 RSI: 0000000000000001 RDI: 
> ffff88c4019b8000
> [    3.074976] RBP: 0000000000000001 R08: 0000000000000002 R09: 
> ffffcd808007b99c
> [    3.074976] R10: ffffcd808007b950 R11: 0000000000000000 R12: 
> 0000000000000001
> [    3.074976] R13: ffff88c4019b80c8 R14: ffff88c401a7e028 R15: 
> ffff88c401a73400
> [    3.074976] FS:  0000000000000000(0000) GS:ffff88d38aad5000(0000) 
> knlGS:0000000000000000
> [    3.074976] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [    3.074976] CR2: 0000000000000021 CR3: 0000000f66222001 CR4: 
> 0000000000770ef0
> [    3.074976] PKRU: 55555554
> [    3.074976] Call Trace:
> [    3.074976]  <TASK>
> [    3.074976]  ? pci_pm_reset+0x39/0x180
> [    3.074976]  pci_init_reset_methods+0x52/0x80
> [    3.074976]  pci_device_add+0x215/0x5d0
> [    3.074976]  pci_scan_single_device+0xa2/0xe0
> [    3.074976]  pci_scan_slot+0x66/0x1c0
> [    3.074976]  ? klist_next+0x145/0x150
> [    3.074976]  pci_scan_child_bus_extend+0x3a/0x290
> [    3.074976]  acpi_pci_root_create+0x236/0x2a0
> [    3.074976]  pci_acpi_scan_root+0x19b/0x1f0
> [    3.074976]  acpi_pci_root_add+0x1a5/0x370
> [    3.074976]  acpi_bus_attach+0x1a8/0x290
> [    3.074976]  ? __pfx_acpi_dev_for_one_check+0x10/0x10
> [    3.074976]  device_for_each_child+0x4b/0x80
> [    3.074976]  acpi_dev_for_each_child+0x28/0x40
> [    3.074976]  ? __pfx_acpi_bus_attach+0x10/0x10
> [    3.074976]  acpi_bus_attach+0x7a/0x290
> [    3.074976]  ? _raw_spin_unlock_irqrestore+0x23/0x40
> [    3.074976]  ? __pfx_acpi_dev_for_one_check+0x10/0x10
> [    3.074976]  device_for_each_child+0x4b/0x80
> [    3.074976]  acpi_dev_for_each_child+0x28/0x40
> [    3.074976]  ? __pfx_acpi_bus_attach+0x10/0x10
> [    3.074976]  acpi_bus_attach+0x7a/0x290
> [    3.074976]  acpi_bus_scan+0x6a/0x1c0
> [    3.074976]  ? __pfx_acpi_init+0x10/0x10
> [    3.074976]  acpi_scan_init+0xdc/0x280
> [    3.074976]  ? __pfx_acpi_init+0x10/0x10
> [    3.074976]  acpi_init+0x218/0x530
> [    3.074976]  do_one_initcall+0x40/0x310
> [    3.074976]  kernel_init_freeable+0x2fe/0x450
> [    3.074976]  ? __pfx_kernel_init+0x10/0x10
> [    3.074976]  kernel_init+0x16/0x1d0
> [    3.074976]  ret_from_fork+0x1ab/0x1e0
> [    3.074976]  ? __pfx_kernel_init+0x10/0x10
> [    3.074976]  ret_from_fork_asm+0x1a/0x30
> [    3.074976]  </TASK>
> [    3.074976] Modules linked in:
> [    3.074976] CR2: 0000000000000021
> [    3.074976] ---[ end trace 0000000000000000 ]---
> [    3.074976] RIP: 0010:pci_reset_bus_function+0xdf/0x160
>
Hi Cedric,

Thanks for pointing this out. I missed that dev->slot could be NULL and 
so the per_func_slot check should be done after the check for 
!dev->slot. I tried this change on top of the patch in an x86_64 VM and 
was able to boot the VM without the oops.

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 70296d3b1cfc..3631f7faa0cf 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -5061,10 +5061,9 @@ static int pci_reset_hotplug_slot(struct 
hotplug_slot *hotplug, bool probe)

  static int pci_dev_reset_slot_function(struct pci_dev *dev, bool probe)
  {
-       if (dev->multifunction && !dev->slot->per_func_slot)
-               return -ENOTTY;
         if (dev->subordinate || !dev->slot ||
-           dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET)
+           dev->dev_flags & PCI_DEV_FLAGS_NO_BUS_RESET ||
+           (dev->multifunction && !dev->slot->per_func_slot))
                 return -ENOTTY;



Thanks
Farhan


  reply	other threads:[~2025-09-16 18:38 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-09-11 18:32 [PATCH v3 00/10] Error recovery for vfio-pci devices on s390x Farhan Ali
2025-09-11 18:32 ` [PATCH v3 01/10] PCI: Avoid saving error values for config space Farhan Ali
2025-09-13  8:27   ` Alex Williamson
2025-09-15 17:15     ` Farhan Ali
2025-09-16 18:09   ` Bjorn Helgaas
2025-09-16 20:00     ` Farhan Ali
2025-09-19 18:17       ` Alex Williamson
2025-09-11 18:32 ` [PATCH v3 02/10] PCI: Add additional checks for flr reset Farhan Ali
2025-09-11 18:33 ` [PATCH v3 03/10] PCI: Allow per function PCI slots Farhan Ali
2025-09-12 12:23   ` Benjamin Block
2025-09-12 17:19     ` Farhan Ali
2025-09-16  6:52   ` Cédric Le Goater
2025-09-16 18:37     ` Farhan Ali [this message]
2025-09-17  6:21       ` Cédric Le Goater
2025-09-17 17:50         ` Farhan Ali
2025-09-11 18:33 ` [PATCH v3 04/10] s390/pci: Add architecture specific resource/bus address translation Farhan Ali
2025-09-17 14:48   ` Niklas Schnelle
2025-09-17 17:22     ` Farhan Ali
2025-09-11 18:33 ` [PATCH v3 05/10] s390/pci: Restore IRQ unconditionally for the zPCI device Farhan Ali
2025-09-15  8:39   ` Niklas Schnelle
2025-09-15 17:42     ` Farhan Ali
2025-09-16 10:59       ` Niklas Schnelle
2025-09-11 18:33 ` [PATCH v3 06/10] s390/pci: Update the logic for detecting passthrough device Farhan Ali
2025-09-15  9:22   ` Niklas Schnelle
2025-09-11 18:33 ` [PATCH v3 07/10] s390/pci: Store PCI error information for passthrough devices Farhan Ali
2025-09-15 11:42   ` Niklas Schnelle
2025-09-15 18:12     ` Farhan Ali
2025-09-16 10:54       ` Niklas Schnelle
2025-09-11 18:33 ` [PATCH v3 08/10] vfio-pci/zdev: Add a device feature for error information Farhan Ali
2025-09-13  9:04   ` Alex Williamson
2025-09-15 18:27     ` Farhan Ali
2025-09-15  6:26   ` Cédric Le Goater
2025-09-15 18:27     ` Farhan Ali
2025-09-11 18:33 ` [PATCH v3 09/10] vfio: Add a reset_done callback for vfio-pci driver Farhan Ali
2025-09-11 18:33 ` [PATCH v3 10/10] vfio: Remove the pcie check for VFIO_PCI_ERR_IRQ_INDEX Farhan Ali

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=e86caff6-8af0-48c9-9058-c1991e23160f@linux.ibm.com \
    --to=alifm@linux.ibm.com \
    --cc=alex.williamson@redhat.com \
    --cc=clg@redhat.com \
    --cc=helgaas@kernel.org \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=linux-s390@vger.kernel.org \
    --cc=mjrosato@linux.ibm.com \
    --cc=schnelle@linux.ibm.com \
    /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