From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from lindbergh.monkeyblade.net ([23.128.96.19]:38542 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S2505954AbgJWM0G (ORCPT ); Fri, 23 Oct 2020 08:26:06 -0400 Received: from mail-wm1-x342.google.com (mail-wm1-x342.google.com [IPv6:2a00:1450:4864:20::342]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 4A230C0613CE for ; Fri, 23 Oct 2020 05:26:06 -0700 (PDT) Received: by mail-wm1-x342.google.com with SMTP id l15so1309611wmi.3 for ; Fri, 23 Oct 2020 05:26:06 -0700 (PDT) Date: Fri, 23 Oct 2020 14:26:02 +0200 From: Daniel Vetter Subject: Re: [PATCH 29/65] s390/pci: Remove races against pte updates Message-ID: <20201023122602.GB401619@phenom.ffwll.local> References: <20201021163242.1458885-1-daniel.vetter@ffwll.ch> <20201023122216.2373294-1-daniel.vetter@ffwll.ch> <20201023122216.2373294-29-daniel.vetter@ffwll.ch> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Disposition: inline Content-Transfer-Encoding: quoted-printable In-Reply-To: <20201023122216.2373294-29-daniel.vetter@ffwll.ch> List-ID: To: DRI Development Cc: Intel Graphics Development , Daniel Vetter , Gerald Schaefer , Daniel Vetter , Jason Gunthorpe , Dan Williams , Kees Cook , Andrew Morton , John Hubbard , =?iso-8859-1?B?Suly9G1l?= Glisse , Jan Kara , linux-mm@kvack.org, linux-arm-kernel@lists.infradead.org, linux-samsung-soc@vger.kernel.org, linux-media@vger.kernel.org, Niklas Schnelle , linux-s390@vger.kernel.org On Fri, Oct 23, 2020 at 02:21:40PM +0200, Daniel Vetter wrote: > Way back it was a reasonable assumptions that iomem mappings never > change the pfn range they point at. But this has changed: >=20 > - gpu drivers dynamically manage their memory nowadays, invalidating > ptes with unmap_mapping_range when buffers get moved >=20 > - contiguous dma allocations have moved from dedicated carvetouts to > cma regions. This means if we miss the unmap the pfn might contain > pagecache or anon memory (well anything allocated with GFP_MOVEABLE) >=20 > - even /dev/mem now invalidates mappings when the kernel requests that > iomem region when CONFIG_IO_STRICT_DEVMEM is set, see commit > 3234ac664a87 ("/dev/mem: Revoke mappings when a driver claims the > region") >=20 > Accessing pfns obtained from ptes without holding all the locks is > therefore no longer a good idea. Fix this. >=20 > Since zpci_memcpy_from|toio seems to not do anything nefarious with > locks we just need to open code get_pfn and follow_pfn and make sure > we drop the locks only after we're done. The write function also needs > the copy_from_user move, since we can't take userspace faults while > holding the mmap sem. >=20 > v2: Move VM_IO | VM_PFNMAP checks around so they keep returning EINVAL > like before (Gerard) >=20 > v3: Polish commit message (Niklas) >=20 > Reviewed-by: Gerald Schaefer > Signed-off-by: Daniel Vetter > Cc: Jason Gunthorpe > Cc: Dan Williams > Cc: Kees Cook > Cc: Andrew Morton > Cc: John Hubbard > Cc: J=EF=BF=BDr=EF=BF=BDme Glisse > Cc: Jan Kara > Cc: linux-mm@kvack.org > Cc: linux-arm-kernel@lists.infradead.org > Cc: linux-samsung-soc@vger.kernel.org > Cc: linux-media@vger.kernel.org > Cc: Niklas Schnelle > Cc: Gerald Schaefer > Cc: linux-s390@vger.kernel.org > Cc: Niklas Schnelle > Signed-off-by: Daniel Vetter Apologies for this random patch bomb, typoed git send-email command :-/ v4 of this will come properly on Monday or so. -Daniel > --- > arch/s390/pci/pci_mmio.c | 98 +++++++++++++++++++++++----------------- > 1 file changed, 57 insertions(+), 41 deletions(-) >=20 > diff --git a/arch/s390/pci/pci_mmio.c b/arch/s390/pci/pci_mmio.c > index 401cf670a243..1a6adbc68ee8 100644 > --- a/arch/s390/pci/pci_mmio.c > +++ b/arch/s390/pci/pci_mmio.c > @@ -119,33 +119,15 @@ static inline int __memcpy_toio_inuser(void __iomem= *dst, > return rc; > } > =20 > -static long get_pfn(unsigned long user_addr, unsigned long access, > - unsigned long *pfn) > -{ > - struct vm_area_struct *vma; > - long ret; > - > - mmap_read_lock(current->mm); > - ret =3D -EINVAL; > - vma =3D find_vma(current->mm, user_addr); > - if (!vma) > - goto out; > - ret =3D -EACCES; > - if (!(vma->vm_flags & access)) > - goto out; > - ret =3D follow_pfn(vma, user_addr, pfn); > -out: > - mmap_read_unlock(current->mm); > - return ret; > -} > - > SYSCALL_DEFINE3(s390_pci_mmio_write, unsigned long, mmio_addr, > const void __user *, user_buffer, size_t, length) > { > u8 local_buf[64]; > void __iomem *io_addr; > void *buf; > - unsigned long pfn; > + struct vm_area_struct *vma; > + pte_t *ptep; > + spinlock_t *ptl; > long ret; > =20 > if (!zpci_is_enabled()) > @@ -158,7 +140,7 @@ SYSCALL_DEFINE3(s390_pci_mmio_write, unsigned long, m= mio_addr, > * We only support write access to MIO capable devices if we are on > * a MIO enabled system. Otherwise we would have to check for every > * address if it is a special ZPCI_ADDR and would have to do > - * a get_pfn() which we don't need for MIO capable devices. Currently > + * a pfn lookup which we don't need for MIO capable devices. Currently > * ISM devices are the only devices without MIO support and there is no > * known need for accessing these from userspace. > */ > @@ -176,21 +158,37 @@ SYSCALL_DEFINE3(s390_pci_mmio_write, unsigned long,= mmio_addr, > } else > buf =3D local_buf; > =20 > - ret =3D get_pfn(mmio_addr, VM_WRITE, &pfn); > + ret =3D -EFAULT; > + if (copy_from_user(buf, user_buffer, length)) > + goto out_free; > + > + mmap_read_lock(current->mm); > + ret =3D -EINVAL; > + vma =3D find_vma(current->mm, mmio_addr); > + if (!vma) > + goto out_unlock_mmap; > + if (!(vma->vm_flags & (VM_IO | VM_PFNMAP))) > + goto out_unlock_mmap; > + ret =3D -EACCES; > + if (!(vma->vm_flags & VM_WRITE)) > + goto out_unlock_mmap; > + > + ret =3D follow_pte_pmd(vma->vm_mm, mmio_addr, NULL, &ptep, NULL, &ptl); > if (ret) > - goto out; > - io_addr =3D (void __iomem *)((pfn << PAGE_SHIFT) | > + goto out_unlock_mmap; > + > + io_addr =3D (void __iomem *)((pte_pfn(*ptep) << PAGE_SHIFT) | > (mmio_addr & ~PAGE_MASK)); > =20 > - ret =3D -EFAULT; > if ((unsigned long) io_addr < ZPCI_IOMAP_ADDR_BASE) > - goto out; > - > - if (copy_from_user(buf, user_buffer, length)) > - goto out; > + goto out_unlock_pt; > =20 > ret =3D zpci_memcpy_toio(io_addr, buf, length); > -out: > +out_unlock_pt: > + pte_unmap_unlock(ptep, ptl); > +out_unlock_mmap: > + mmap_read_unlock(current->mm); > +out_free: > if (buf !=3D local_buf) > kfree(buf); > return ret; > @@ -274,7 +272,9 @@ SYSCALL_DEFINE3(s390_pci_mmio_read, unsigned long, mm= io_addr, > u8 local_buf[64]; > void __iomem *io_addr; > void *buf; > - unsigned long pfn; > + struct vm_area_struct *vma; > + pte_t *ptep; > + spinlock_t *ptl; > long ret; > =20 > if (!zpci_is_enabled()) > @@ -287,7 +287,7 @@ SYSCALL_DEFINE3(s390_pci_mmio_read, unsigned long, mm= io_addr, > * We only support read access to MIO capable devices if we are on > * a MIO enabled system. Otherwise we would have to check for every > * address if it is a special ZPCI_ADDR and would have to do > - * a get_pfn() which we don't need for MIO capable devices. Currently > + * a pfn lookup which we don't need for MIO capable devices. Currently > * ISM devices are the only devices without MIO support and there is no > * known need for accessing these from userspace. > */ > @@ -306,22 +306,38 @@ SYSCALL_DEFINE3(s390_pci_mmio_read, unsigned long, = mmio_addr, > buf =3D local_buf; > } > =20 > - ret =3D get_pfn(mmio_addr, VM_READ, &pfn); > + mmap_read_lock(current->mm); > + ret =3D -EINVAL; > + vma =3D find_vma(current->mm, mmio_addr); > + if (!vma) > + goto out_unlock_mmap; > + if (!(vma->vm_flags & (VM_IO | VM_PFNMAP))) > + goto out_unlock_mmap; > + ret =3D -EACCES; > + if (!(vma->vm_flags & VM_WRITE)) > + goto out_unlock_mmap; > + > + ret =3D follow_pte_pmd(vma->vm_mm, mmio_addr, NULL, &ptep, NULL, &ptl); > if (ret) > - goto out; > - io_addr =3D (void __iomem *)((pfn << PAGE_SHIFT) | (mmio_addr & ~PAGE_M= ASK)); > + goto out_unlock_mmap; > + > + io_addr =3D (void __iomem *)((pte_pfn(*ptep) << PAGE_SHIFT) | > + (mmio_addr & ~PAGE_MASK)); > =20 > if ((unsigned long) io_addr < ZPCI_IOMAP_ADDR_BASE) { > ret =3D -EFAULT; > - goto out; > + goto out_unlock_pt; > } > ret =3D zpci_memcpy_fromio(buf, io_addr, length); > - if (ret) > - goto out; > - if (copy_to_user(user_buffer, buf, length)) > + > +out_unlock_pt: > + pte_unmap_unlock(ptep, ptl); > +out_unlock_mmap: > + mmap_read_unlock(current->mm); > + > + if (!ret && copy_to_user(user_buffer, buf, length)) > ret =3D -EFAULT; > =20 > -out: > if (buf !=3D local_buf) > kfree(buf); > return ret; > --=20 > 2.28.0 >=20 --=20 Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch