Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
From:   Anatol Pomozov <anatol.pomozov@gmail.com>
Date:   Thu, 5 Dec 2019 13:18:52 -0800
Message-ID: <CAOMFOmXU8RoeAOEjP=gkkiZdPS8Ok_wzXcgUZxL11u1zMSEtMQ@mail.gmail.com>
Subject: dmesg spammed with "caller memremap+0x64/0x160 mapping multiple BARs"
To:     LKML <linux-kernel@vger.kernel.org>
Cc:     avid@redhat.com, hch@lst.de, namit@vmware.com, tglx@linutronix.de,
        delvare@suse.de, rrichter@marvell.com, rppt@linux.vnet.ibm.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi folks

We have several servers that run on kernel 4.14.12 and their logs
spammed with following messages:

 [20633911.311910] caller memremap+0x64/0x160 mapping multiple BARs
 [20633911.318577] resource sanity check: requesting [mem
0x000eb570-0x000edc34], which spans more than PCI Bus 0000:00 [mem
0x000e8000-0x000ebfff window]
 [20633911.333650] caller memremap+0x64/0x160 mapping multiple BARs
 [20633911.340412] resource sanity check: requesting [mem
0x000eb570-0x000edc34], which spans more than PCI Bus 0000:00 [mem
0x000e8000-0x000ebfff window]
 [20633911.355481] caller memremap+0x64/0x160 mapping multiple BARs

After some debugging I've found that the warning is printed when I open the file
/sys/firmware/dmi/entries/*/raw The stacktrace is

[   14.839284] Call Trace:
[   14.839289]  dump_stack+0x46/0x59
[   14.839293]  iomem_map_sanity_check+0xb8/0xc0
[   14.839296]  __ioremap_caller+0x225/0x2f0
[   14.839302]  memremap+0x113/0x170
[   14.856703]  ? dmi_name_in_serial+0x30/0x30
[   14.856705]  dmi_walk+0x30/0x70
[   14.856706]  find_dmi_entry+0x46/0x80
[   14.856708]  ? dmi_entry_free+0x10/0x10
[   14.856710]  dmi_entry_raw_read+0x3a/0x60
[   14.856712]  kernfs_fop_read+0xa1/0x170
[   14.856714]  __vfs_read+0x36/0x170
[   14.856717]  vfs_read+0x89/0x130
[   14.856719]  SyS_read+0x52/0xc0
[   14.856721]  do_syscall_64+0x60/0x110
[   14.856723]  entry_SYSCALL_64_after_hwframe+0x3d/0xa2
[   14.856725] RIP: 0033:0x7f60aceedf70

As it turned out the root of the problem is BIOS region mapping. Here
are ACPI mappings:

000e0000-000fffff : Reserved
  000e0000-000e3fff : PCI Bus 0000:00
  000e4000-000e7fff : PCI Bus 0000:00
  000e8000-000ebfff : PCI Bus 0000:00
  000ec000-000effff : PCI Bus 0000:00
  000f0000-000fffff : PCI Bus 0000:00

Or the same mapping from IASL tool:

DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed,
Cacheable, ReadWrite,
    0x00000000,         // Granularity
    0x000E8000,         // Range Minimum
    0x000EBFFF,         // Range Maximum
    0x00000000,         // Translation Offset
    0x00004000,         // Length
    ,, , AddressRangeMemory, TypeStatic)
DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed,
Cacheable, ReadWrite,
    0x00000000,         // Granularity
    0x000EC000,         // Range Minimum
    0x000EFFFF,         // Range Maximum
    0x00000000,         // Translation Offset
    0x00004000,         // Length
    ,, , AddressRangeMemory, TypeStatic)


SMBIOS table is located at 0x000eb570-0x000edc34 address range. When
find_dmi_entry() tries to map SMBIOS to kernel address space the
warning above is printed.


So I have a few questions here:
 - What exactly kernel tries to tell us here? Why mapping SMBIOS over
multiple PCI mappings is bad. Sorry I am not familiar with this
codepath and I am trying to understand the issue.
 - Is the warning really bad? The code works just fine despite the
warnings. Can the warning be turned into info/debug()?
 - Is the mapping need to be fixed in BIOS? What other possible
problems can happen due to such BIOS mapping misconfiguration?
 - Why SMBIOS table mapped for every sysfs file read? Would it make
sense to map SMBIOS only once at the kernel boot?
 - Does it make sense to ratelimit the warning? I will be glad to send
the path for it.