[PATCH 1/1] PCI: Do not add hotplug reservation multiple times

[PATCH 1/1] PCI: Do not add hotplug reservation multiple times

[Ilpo Järvinen]

In nested topologies, hotplug reservations gets added multiple times
into the bridge window higher up in the hierarchy. Adding reservation
for intermediate level bridges does not seem very useful because the
hotplug is going to add device at a leaf.

Accounting the hoptlug reservation multiple times results in larger
than expected bridge window size that may lead to assignment failures
as show in this log:

pci_bus 0000:0a: root bus resource [mem 0x10a00000-0x10c00fff window]
pci 0000:0a:00.0: BAR 0 [mem 0x10c00000-0x10c00fff]
pci 0000:0a:00.0:   bridge window [mem 0x10a00000-0x10bfffff]
pci 0000:0b:00.0:   bridge window [mem 0x10a00000-0x10bfffff]
pci 0000:0c:02.0:   bridge window [mem 0x10a00000-0x10bfffff]
pci 0000:0c:02.0: bridge window [mem 0x00100000-0x000fffff 64bit pref] to [bus 0d] add_size 200000 add_align 100000
pci 0000:0c:02.0: bridge window [mem 0x00100000-0x000fffff] to [bus 0d] add_size 200000 add_align 100000
pci 0000:0b:00.0: bridge window [mem 0x00100000-0x000fffff 64bit pref] to [bus 0c-0d] add_size 200000 add_align 100000
pci 0000:0b:00.0: bridge window [mem 0x00100000-0x000fffff] to [bus 0c-0d] add_size 200000 add_align 100000
pci 0000:0a:00.0: bridge window [mem 0x00100000-0x000fffff 64bit pref] to [bus 0b-0d] add_size 400000 add_align 100000
pci 0000:0a:00.0: bridge window [mem 0x00100000-0x000fffff] to [bus 0b-0d] add_size 400000 add_align 100000
pci 0000:0a:00.0: bridge window [mem size 0x00400000]: can't assign; no space
pci 0000:0a:00.0: bridge window [mem size 0x00400000]: failed to assign
pci 0000:0a:00.0: bridge window [mem size 0x00400000 64bit pref]: can't assign; no space
pci 0000:0a:00.0: bridge window [mem size 0x00400000 64bit pref]: failed to assign

The problem stems from calculate_memsize() that calculates size first
and then adds childen size:
	size = max(0, 2M) + 2M

Alter the logic to first account for the children size before applying
the hotplug reservation to not add hotplug reservation multiple times
on different levels of a nested topology. As a result, the size
calculation becomes:
	size = max(0 + 2M, 2M)

Reported-by: Eric Auger <eauger@redhat.com>
Link: https://lore.kernel.org/linux-pci/f23946f2-06ac-4607-8f2c-3ffbc52b627a@redhat.com/
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
---
 drivers/pci/setup-bus.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c
index c0a949f2c995..e8c94aa1d3c1 100644
--- a/drivers/pci/setup-bus.c
+++ b/drivers/pci/setup-bus.c
@@ -1021,7 +1021,7 @@ static resource_size_t calculate_memsize(resource_size_t size,
 					 resource_size_t children_add_size,
 					 resource_size_t align)
 {
-	size = max(size, min_size) + children_add_size;
+	size = max(size + children_add_size, min_size);
 	return ALIGN(size, align);
 }
 

base-commit: dc59e4fea9d83f03bad6bddf3fa2e52491777482
-- 
2.47.3

Re: [PATCH 1/1] PCI: Do not add hotplug reservation multiple times

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] The patch fixes nested hotplug reservation accumulation for memory windows in `calculate_memsize()`, but misses the identical buggy logic in `calculate_iosize()`, leading to I/O space exhaustion.
--

--- Patch [1]: [PATCH 1/1] PCI: Do not add hotplug reservation multiple times ---
commit 3c4b4a62adcb92793a7c17214021e8b637be1016
Author: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>

PCI: Do not add hotplug reservation multiple times

In nested topologies, hotplug reservations gets added multiple times
into the bridge window higher up in the hierarchy. Adding reservation
for intermediate level bridges does not seem very useful because the
hotplug is going to add device at a leaf.

Link: https://lore.kernel.org/linux-pci/f23946f2-06ac-4607-8f2c-3ffbc52b627a@redhat.com/

> diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c
> index c0a949f2c9956..e8c94aa1d3c12 100644
> --- a/drivers/pci/setup-bus.c
> +++ b/drivers/pci/setup-bus.c
> @@ -1021,7 +1021,7 @@ static resource_size_t calculate_memsize(resource_size_t size,
>  					 resource_size_t children_add_size,
>  					 resource_size_t align)
>  {
> -	size = max(size, min_size) + children_add_size;
> +	size = max(size + children_add_size, min_size);
>  	return ALIGN(size, align);
>  }

[Severity: High]
This isn't a bug introduced by this patch, but does calculate_iosize() need a
similar fix to avoid I/O space exhaustion in nested topologies?

Just above calculate_memsize(), calculate_iosize() retains the identical
logic for I/O windows:

drivers/pci/setup-bus.c:calculate_iosize() {
    ...
    size = max(size, add_size) + children_add_size;
    return ALIGN(max(size, old_size), align);
}

When allocating I/O windows for nested hotplug bridges, pbus_size_io() passes
both the accumulated children I/O sizes (children_add_size) and the hotplug
reservation (add_size) to calculate_iosize().

Because the logic is exactly the same as the unpatched calculate_memsize(),
this appears to still unconditionally add an extra 4KB (due to bridge window
alignment) at each nested bridge level. Can this linearly inflate the required
space and quickly exhaust the limited 64KB I/O address space on systems with
deeply nested hotplug PCIe bridges (like Thunderbolt docks)?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260720162416.5771-1-ilpo.jarvinen@linux.intel.com?part=1

Re: [PATCH 1/1] PCI: Do not add hotplug reservation multiple times

[Ilpo Järvinen]

[-- Attachment #1: Type: text/plain, Size: 2956 bytes --]

On Mon, 20 Jul 2026, sashiko-bot@kernel.org wrote:

> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> 
> Pre-existing issues:
> - [High] The patch fixes nested hotplug reservation accumulation for memory windows in `calculate_memsize()`, but misses the identical buggy logic in `calculate_iosize()`, leading to I/O space exhaustion.
> --
> 
> --- Patch [1]: [PATCH 1/1] PCI: Do not add hotplug reservation multiple times ---
> commit 3c4b4a62adcb92793a7c17214021e8b637be1016
> Author: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
> 
> PCI: Do not add hotplug reservation multiple times
> 
> In nested topologies, hotplug reservations gets added multiple times
> into the bridge window higher up in the hierarchy. Adding reservation
> for intermediate level bridges does not seem very useful because the
> hotplug is going to add device at a leaf.
> 
> Link: https://lore.kernel.org/linux-pci/f23946f2-06ac-4607-8f2c-3ffbc52b627a@redhat.com/
> 
> > diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c
> > index c0a949f2c9956..e8c94aa1d3c12 100644
> > --- a/drivers/pci/setup-bus.c
> > +++ b/drivers/pci/setup-bus.c
> > @@ -1021,7 +1021,7 @@ static resource_size_t calculate_memsize(resource_size_t size,
> >  					 resource_size_t children_add_size,
> >  					 resource_size_t align)
> >  {
> > -	size = max(size, min_size) + children_add_size;
> > +	size = max(size + children_add_size, min_size);
> >  	return ALIGN(size, align);
> >  }
> 
> [Severity: High]
> This isn't a bug introduced by this patch, but does calculate_iosize() need a
> similar fix to avoid I/O space exhaustion in nested topologies?
> 
> Just above calculate_memsize(), calculate_iosize() retains the identical
> logic for I/O windows:
>
> drivers/pci/setup-bus.c:calculate_iosize() {
>     ...
>     size = max(size, add_size) + children_add_size;
>     return ALIGN(max(size, old_size), align);
> }
> 
> When allocating I/O windows for nested hotplug bridges, pbus_size_io() passes
> both the accumulated children I/O sizes (children_add_size) and the hotplug
> reservation (add_size) to calculate_iosize().
> 
> Because the logic is exactly the same as the unpatched calculate_memsize(),
> this appears to still unconditionally add an extra 4KB (due to bridge window
> alignment) at each nested bridge level. Can this linearly inflate the required
> space and quickly exhaust the limited 64KB I/O address space on systems with
> deeply nested hotplug PCIe bridges (like Thunderbolt docks)?

For the record, I intentionally left the io side untouched.

If somebody feels I should make the same change there, I can of course add 
it there.

IMO io side would be better served with defaulting hotplug allocation to 
zero, big systems keep running out of io space because of gazillions empty 
reservations even when nesting is not in the picture.


-- 
 i.