| 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
| ||
|
Date: Wed, 06 Mar 2013 11:19:10 +0000
From: Matt Fleming <matt@...sole-pimps.org>
To: joeyli <jlee@...e.com>
Cc: linux-efi@...r.kernel.org, linux-kernel@...r.kernel.org,
Michael Schroeder <mls@...e.com>,
Josh Boyer <jwboyer@...hat.com>,
Peter Jones <pjones@...hat.com>,
Matthew Garrett <mjg59@...f.ucam.org>,
Frederic Crozat <fcrozat@...e.com>
Subject: Re: [PATCH] efivarfs: fix abnormal GUID in variable name by using
strcpy to replace null with dash
On Wed, 2013-03-06 at 15:34 +0800, joeyli wrote:
> Hi Matt,
>
> 於 六，2013-03-02 於 07:41 +0800，joeyli 提到：
> > 於 五，2013-03-01 於 16:31 +0000，Matt Fleming 提到：
> > > On Fri, 2013-03-01 at 15:17 +0000, Matt Fleming wrote:
> > > > On Fri, 2013-03-01 at 11:20 +0800, Lee, Chun-Yi wrote:
> > > > > From: Michael Schroeder <mls@...e.com>
> > > > >
> > > > > On HP z220 system (firmware version 1.54), some EFI variables are incorrectly
> > > > > named :
> > > > >
> > > > > ls -d /sys/firmware/efi/vars/*8be4d* | grep -v -- -8be returns
> > > > > /sys/firmware/efi/vars/dbxDefault-pport8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > > /sys/firmware/efi/vars/KEKDefault-pport8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > > /sys/firmware/efi/vars/SecureBoot-pport8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > > /sys/firmware/efi/vars/SetupMode-Information8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > >
> > > > > That causes by the following statement in efivar_create_sysfs_entry function:
> > > > >
> > > > > *(short_name + strlen(short_name)) = '-';
> > > > > efi_guid_unparse(vendor_guid, short_name + strlen(short_name));
> > > > >
> > > > > The trailing \0 is overwritten with '-', but the next char doesn't seem to be a \0
> > > > > as well for HP. So, the second strlen return the point of next '\0', causes there
> > > > > have garbage string attached before GUID.
> > > > >
> > > > > Tested on On HP z220.
> > > >
> > > > What's more likely happening here is that GetNextVariable() is broken on
> > > > this HP firmware and variable_name_size is too big for the given
> > > > variable in variable_name. We've seen other reports of similar bugs,
> > > >
> > > > https://bugzilla.kernel.org/show_bug.cgi?id=47631
> > > >
> > > >
> > > > Could someone try this patch against Linus' tree?
> > >
> > > Urgh, and here's a version that isn't utterly, utterly broken...
> >
> > Thanks for Matt's patch, we will try it!
> >
> > Joey Lee
>
> Frederic confirmed your patch works to him for fix issue on HP machine.
>
> Tested-by: Frederic Crozat <fcrozat@...e.com>
>
> But, I suggest we just use the length with NULL in variable_name but not
> VariableNameSize that was returned by GetNextVariableName.
>
> My thinking is following...
>
> >
> > >
> > > ---
> > >
> > > >From 4b2ef72bca72039717efe4570ec858a86d565b34 Mon Sep 17 00:00:00 2001
> > > From: Matt Fleming <matt.fleming@...el.com>
> > > Date: Fri, 1 Mar 2013 14:49:12 +0000
> > > Subject: [PATCH] efivars: Sanitise string length returned by
> > > GetNextVariableName()
> > >
> > > Some buggy firmware implementations return a string length from
> > > GetNextVariableName() that is actually larger than the string in
> > > 'variable_name', as Michael Schroeder writes,
> > >
> > > > On HP z220 system (firmware version 1.54), some EFI variables are
> > > > incorrectly named :
> > > >
> > > > ls -d /sys/firmware/efi/vars/*8be4d* | grep -v -- -8be returns
> > > > /sys/firmware/efi/vars/dbxDefault-pport8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > /sys/firmware/efi/vars/KEKDefault-pport8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > /sys/firmware/efi/vars/SecureBoot-pport8be4df61-93ca-11d2-aa0d-00e098032b8c
> > > > /sys/firmware/efi/vars/SetupMode-Information8be4df61-93ca-11d2-aa0d-00e098032b8c
> > >
> > > Since 'variable_name' is a string, we can validate its size by
> > > searching for the terminating NULL character.
> > >
> > > Reported-by: Frederic Crozat <fcrozat@...e.com>
> > > Cc: Matthew Garrett <mjg59@...f.ucam.org>
> > > Cc: Josh Boyer <jwboyer@...hat.com>
> > > Cc: Michael Schroeder <mls@...e.com>
> > > Cc: Lee, Chun-Yi <jlee@...e.com>
> > > Cc: Lingzhu Xiang <lxiang@...hat.com>
> > > Signed-off-by: Matt Fleming <matt.fleming@...el.com>
> > > ---
> > > drivers/firmware/efivars.c | 30 ++++++++++++++++++++++++++++++
> > > 1 file changed, 30 insertions(+)
> > >
> > > diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
> > > index 7320bf8..ab477b8 100644
> > > --- a/drivers/firmware/efivars.c
> > > +++ b/drivers/firmware/efivars.c
> > > @@ -1895,6 +1895,33 @@ void unregister_efivars(struct efivars *efivars)
> > > }
> > > EXPORT_SYMBOL_GPL(unregister_efivars);
> > >
> > > +/*
> > > + * Sanity check size of a variable name.
> > > + */
> > > +static unsigned long sanity_check_size(efi_char16_t *variable_name,
> > > + unsigned long variable_name_size)
> > > +{
> > > + unsigned long len;
> > > + efi_char16_t c;
> > > +
> > > + /*
> > > + * The variable name is, by definition, a NULL-terminated
> > > + * string, so make absolutely sure that variable_name_size is
> > > + * the value we expect it to be. If not, return the real size.
> > > + */
> > > + for (len = 2; len <= variable_name_size; len += sizeof(c)) {
> > > + c = variable_name[(len / sizeof(c)) - 1];
> > > + if (!c)
> > > + break;
> > > + }
> > > +
>
> We just direct return the len here but don't need compare with
> variable_name_size.
>
> return len;
>
> > > +
> > > + if (len != variable_name_size)
> > > + printk(KERN_WARNING "efivars: bogus variable_name_size: %lu %lu\n", len, variable_name_size);
> > > +
> > > + return min(len, variable_name_size);
> > > +}
> > > +
>
> In UEFI 2.3.1 spec:
>
> VariableNameSize The size of the VariableName buffer
>
> Note that if EFI_BUFFER_TOO_SMALL is returned, the VariableName buffer
> was too small for the next variable. When such an error occurs, the
> VariableNameSize is updated to reflect the size of buffer needed. In all
> cases when calling GetNextVariableName() the VariableNameSize must not
> exceed the actual buffer size that was allocated for VariableName.
>
>
> Per spec, VariableNameSize will updated to 'the size of buffer needed'
> when EFI_BUFFER_TOO_SMALL, but spec doesn't mention VariableNameSize
> will also updated when EFI_SUCCESS. That means 'VariableNameSize = 1024'
> when EFI_SUCCESS is not a bogus value.
>
> > > int register_efivars(struct efivars *efivars,
> > > const struct efivar_operations *ops,
> > > struct kobject *parent_kobj)
> > > @@ -1941,8 +1968,11 @@ int register_efivars(struct efivars *efivars,
> > > status = ops->get_next_variable(&variable_name_size,
> > > variable_name,
> > > &vendor_guid);
> > > +
> > > switch (status) {
> > > case EFI_SUCCESS:
> > > + variable_name_size = sanity_check_size(variable_name,
> > > + variable_name_size);
> > > efivar_create_sysfs_entry(efivars,
> > > variable_name_size,
> > > variable_name,
> >
>
> Due to variable_name_size could be 1024, so I suggest direct feed the
> length of variable_name to efivar_create_sysfs_entry since we are need
> to count the length.
>
> This is a simply diff for my think:
>
>
> diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
> index 7320bf8..99a4f9f 100644
> --- a/drivers/firmware/efivars.c
> +++ b/drivers/firmware/efivars.c
> @@ -1895,6 +1895,21 @@ void unregister_efivars(struct efivars *efivars)
> }
> EXPORT_SYMBOL_GPL(unregister_efivars);
>
> +static unsigned long variable_name_length(efi_char16_t *variable_name)
> +{
> + unsigned long len;
> + efi_char16_t c;
> +
> + len = 2;
> + do {
> + c = variable_name[len / sizeof(c) - 1];
> + if (c)
> + len += sizeof(c);
> + } while (c);
> +
> + return len;
> +}
> +
> int register_efivars(struct efivars *efivars,
> const struct efivar_operations *ops,
> struct kobject *parent_kobj)
> @@ -1944,7 +1959,7 @@ int register_efivars(struct efivars *efivars,
> switch (status) {
> case EFI_SUCCESS:
> efivar_create_sysfs_entry(efivars,
> - variable_name_size,
> + variable_name_length(variable_name),
> variable_name,
> &vendor_guid);
> break;
>
Hmm.. the reason I didn't implement the patch this way is because I do
think it's important to make sure we don't go out of bounds looking for
the terminating NULL, i.e. you need a 'len < variable_name_size' check
somewhere.
Care to update and resend your patch, ensuring we don't inspect more
than variable_name_size characters?
Also, which machine did you see this behaviour on?
--
Matt Fleming, Intel Open Source Technology Center
--
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/
Powered by blists - more mailing lists