linux-kernel - Re: [PATCH v12 0/10] Support Write-Through mapping on x86

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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1433270808.23540.155.camel@misato.fc.hp.com>
Date:	Tue, 02 Jun 2015 12:46:48 -0600
From:	Toshi Kani <toshi.kani@...com>
To:	"Luis R. Rodriguez" <mcgrof@...e.com>
Cc:	bp@...en8.de, hpa@...or.com, tglx@...utronix.de, mingo@...hat.com,
	akpm@...ux-foundation.org, arnd@...db.de, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org, x86@...nel.org,
	linux-nvdimm@...ts.01.org, jgross@...e.com,
	stefan.bader@...onical.com, luto@...capital.net, hmh@....eng.br,
	yigal@...xistor.com, konrad.wilk@...cle.com, Elliott@...com,
	hch@....de
Subject: Re: [PATCH v12 0/10] Support Write-Through mapping on x86

On Tue, 2015-06-02 at 18:21 +0200, Luis R. Rodriguez wrote:
> On Mon, Jun 01, 2015 at 01:36:23PM -0600, Toshi Kani wrote:
> > This patchset adds support of Write-Through (WT) mapping on x86.
> > The study below shows that using WT mapping may be useful for
> > non-volatile memory.
> > 
> > http://www.hpl.hp.com/techreports/2012/HPL-2012-236.pdf
> > 
> > The patchset consists of the following changes.
> >  - Patch 1/10 to 6/10 add ioremap_wt()
> >  - Patch 7/10 adds pgprot_writethrough()
> >  - Patch 8/10 to 9/10 add set_memory_wt()
> >  - Patch 10/10 changes the pmem driver to call ioremap_wt()
> > 
> > All new/modified interfaces have been tested.
> > 
> > The patchset is based on:
> > git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git#tip-mm-2
> 
> While at it can you also look at:
>
> mcgrof@...on ~/linux-next (git::master)$ git grep -4 "writethrough" drivers/infiniband/

Thanks for checking this.  The inifiniband driver uses WT mappings on
powerpc without proper WT interfaces defined.  They can be cleaned up by
a separate patch series to support WT on powerpc in the same way after
this patchset (support WT on x86) is settled.

> drivers/infiniband/hw/ipath/ipath_driver.c-
> drivers/infiniband/hw/ipath/ipath_driver.c-     dd->ipath_pcirev = pdev->revision;
> drivers/infiniband/hw/ipath/ipath_driver.c-
> drivers/infiniband/hw/ipath/ipath_driver.c-#if defined(__powerpc__)
> drivers/infiniband/hw/ipath/ipath_driver.c:     /* There isn't a generic way to specify writethrough mappings */
> drivers/infiniband/hw/ipath/ipath_driver.c-     dd->ipath_kregbase = __ioremap(addr, len,
> drivers/infiniband/hw/ipath/ipath_driver.c-             (_PAGE_NO_CACHE|_PAGE_WRITETHRU));
> drivers/infiniband/hw/ipath/ipath_driver.c-#else
> drivers/infiniband/hw/ipath/ipath_driver.c-     dd->ipath_kregbase = ioremap_nocache(addr, len);

Once powerpc defines ioremap_wt(), this driver can simply call
ioremap_wt() without the ifdef.  ioremap_wt() is defined as
ioremap_nocache() on the architectures without WT support.

> drivers/infiniband/hw/ipath/ipath_file_ops.c-
> drivers/infiniband/hw/ipath/ipath_file_ops.c-   phys = dd->ipath_physaddr + piobufs;
> drivers/infiniband/hw/ipath/ipath_file_ops.c-
> drivers/infiniband/hw/ipath/ipath_file_ops.c-#if defined(__powerpc__)
> drivers/infiniband/hw/ipath/ipath_file_ops.c:   /* There isn't a generic way to specify writethrough mappings */
> drivers/infiniband/hw/ipath/ipath_file_ops.c-   pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
> drivers/infiniband/hw/ipath/ipath_file_ops.c-   pgprot_val(vma->vm_page_prot) |= _PAGE_WRITETHRU;
> drivers/infiniband/hw/ipath/ipath_file_ops.c-   pgprot_val(vma->vm_page_prot) &= ~_PAGE_GUARDED;
> drivers/infiniband/hw/ipath/ipath_file_ops.c-#endif

Once powerpc defines pgprot_writethrough(), this driver can use it
without the ifdef.  pgprot_writethrough() is defined as
pgprot_noncached() on the architectures without WT support.

> drivers/infiniband/hw/qib/qib_file_ops.c-
> drivers/infiniband/hw/qib/qib_file_ops.c-       phys = dd->physaddr + piobufs;
> drivers/infiniband/hw/qib/qib_file_ops.c-
> drivers/infiniband/hw/qib/qib_file_ops.c-#if defined(__powerpc__)
> drivers/infiniband/hw/qib/qib_file_ops.c:       /* There isn't a generic way to specify writethrough mappings */
> drivers/infiniband/hw/qib/qib_file_ops.c-       pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
> drivers/infiniband/hw/qib/qib_file_ops.c-       pgprot_val(vma->vm_page_prot) |= _PAGE_WRITETHRU;
> drivers/infiniband/hw/qib/qib_file_ops.c-       pgprot_val(vma->vm_page_prot) &= ~_PAGE_GUARDED;
> drivers/infiniband/hw/qib/qib_file_ops.c-#endif

Same as above.

> drivers/infiniband/hw/qib/qib_pcie.c-   addr = pci_resource_start(pdev, 0);
> drivers/infiniband/hw/qib/qib_pcie.c-   len = pci_resource_len(pdev, 0);
> drivers/infiniband/hw/qib/qib_pcie.c-
> drivers/infiniband/hw/qib/qib_pcie.c-#if defined(__powerpc__)
> drivers/infiniband/hw/qib/qib_pcie.c:   /* There isn't a generic way to specify writethrough mappings */
> drivers/infiniband/hw/qib/qib_pcie.c-   dd->kregbase = __ioremap(addr, len, _PAGE_NO_CACHE | _PAGE_WRITETHRU);
> drivers/infiniband/hw/qib/qib_pcie.c-#else
> drivers/infiniband/hw/qib/qib_pcie.c-   dd->kregbase = ioremap_nocache(addr, len);
> drivers/infiniband/hw/qib/qib_pcie.c-#endif

Same as ipath_driver.c.

Thanks,
-Toshi

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/