lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20220825083713.4glfivegmodluiun@pali>
Date:   Thu, 25 Aug 2022 10:37:13 +0200
From:   Pali Rohár <pali@...nel.org>
To:     Michael Ellerman <mpe@...erman.id.au>
Cc:     Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        linuxppc-dev@...ts.ozlabs.org, linux-pci@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] powerpc/pci: Enable PCI domains in /proc when PCI bus
 numbers are not unique

On Thursday 25 August 2022 17:49:28 Michael Ellerman wrote:
> Pali Rohár <pali@...nel.org> writes:
> > On 32-bit powerpc systems with more PCIe controllers and more PCI domains,
> > where on more PCI domains are same PCI numbers, when kernel is compiled
> > with CONFIG_PROC_FS=y and CONFIG_PPC_PCI_BUS_NUM_DOMAIN_DEPENDENT=y
> > options, kernel prints "proc_dir_entry 'pci/01' already registered" error
> > message.
> 
> Thanks, I'll pick this up.
> 
> > This regression started appearing after commit 566356813082 ("powerpc/pci:
> > Add config option for using all 256 PCI buses") in case in each mPCIe slot
> > is connected PCIe card and therefore PCI bus 1 is populated in for every
> > PCIe controller / PCI domain.
> >
> > The reason is that PCI procfs code expects that when PCI bus numbers are
> > not unique across all PCI domains, function pci_proc_domain() returns true
> > for domain dependent buses.
> >
> > Fix this issue by setting PCI_ENABLE_PROC_DOMAINS and PCI_COMPAT_DOMAIN_0
> > flags for 32-bit powerpc code when CONFIG_PPC_PCI_BUS_NUM_DOMAIN_DEPENDENT
> > is enabled. Same approach is already implemented for 64-bit powerpc code
> > (where PCI bus numbers are always domain dependent).
> 
> We also have the same in ppc4xx_pci_find_bridges().
> 
> And if we can eventually make CONFIG_PPC_PCI_BUS_NUM_DOMAIN_DEPENDENT
> the standard behaviour on 32-bit then everything would behave the same
> and we could simplify pci_proc_domain() to match what other arches do.

I sent two patches which do another steps to achieve it:
https://lore.kernel.org/linuxppc-dev/20220817163927.24453-1-pali@kernel.org/t/#u

Main blocker is pci-OF-bus-map which is in direct conflict with
CONFIG_PPC_PCI_BUS_NUM_DOMAIN_DEPENDENT and which used on chrp and pmac.
And I have no idea if pci-OF-bus-map is still needed or not.

> cheers
> 
> 
> > Fixes: 566356813082 ("powerpc/pci: Add config option for using all 256 PCI buses")
> > Signed-off-by: Pali Rohár <pali@...nel.org>
> > ---
> >  arch/powerpc/kernel/pci_32.c | 9 +++++++++
> >  1 file changed, 9 insertions(+)
> >
> > diff --git a/arch/powerpc/kernel/pci_32.c b/arch/powerpc/kernel/pci_32.c
> > index ffc4e1928c80..8acbc9592ebb 100644
> > --- a/arch/powerpc/kernel/pci_32.c
> > +++ b/arch/powerpc/kernel/pci_32.c
> > @@ -249,6 +249,15 @@ static int __init pcibios_init(void)
> >  
> >  	printk(KERN_INFO "PCI: Probing PCI hardware\n");
> >  
> > +#ifdef CONFIG_PPC_PCI_BUS_NUM_DOMAIN_DEPENDENT
> > +	/*
> > +	 * Enable PCI domains in /proc when PCI bus numbers are not unique
> > +	 * across all PCI domains to prevent conflicts. And keep PCI domain 0
> > +	 * backward compatible in /proc for video cards.
> > +	 */
> > +	pci_add_flags(PCI_ENABLE_PROC_DOMAINS | PCI_COMPAT_DOMAIN_0);
> > +#endif
> > +
> >  	if (pci_has_flag(PCI_REASSIGN_ALL_BUS))
> >  		pci_assign_all_buses = 1;
> >  
> > -- 
> > 2.20.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ