block: Add option to use driver whitelist even in tools
PATCH 1/1 block: Add option to use driver whitelist even in tools
Currently, the block driver whitelists are only applied for the system emulator. All other binaries still give unrestricted access to all block drivers. There are use cases where this made sense because the main concern was avoiding customers running VMs on less optimised block drivers and getting bad performance. Allowing the same image format e.g. as a target for 'qemu-img convert' is not a problem then.
However, if the concern is the supportability of the driver in general, either in full or when used read-write, not applying the list driver whitelist in tools doesn't help - especially since qemu-nbd and qemu-storage-daemon now give access to more or less the same operations in block drivers as running a system emulator.
In order to address this, introduce a new configure option that enforces the driver whitelist in all binaries.
--- configure | 14 ++++++++++++-- block.c | 3 +++ meson.build | 1 + 3 files changed, 16 insertions(+), 2 deletions(-)
configure +12 -2
@@ -243,6 +243,7 @@ cross_prefix="" audio_drv_list="" block_drv_rw_whitelist="" block_drv_ro_whitelist="" +block_drv_whitelist_tools="no" host_cc="cc" audio_win_int="" libs_qga=""
@@ -1003,6 +1004,10 @@ for opt do ;; --block-drv-ro-whitelist=*) block_drv_ro_whitelist=$(echo "$optarg" | sed -e 's/,/ /g') ;; + --enable-block-drv-whitelist-in-tools) block_drv_whitelist_tools="yes" + ;; + --disable-block-drv-whitelist-in-tools) block_drv_whitelist_tools="no" + ;; --enable-debug-tcg) debug_tcg="yes" ;; --disable-debug-tcg) debug_tcg="no"
@@ -1776,10 +1781,12 @@ Advanced options (experts only): --block-drv-whitelist=L Same as --block-drv-rw-whitelist=L --block-drv-rw-whitelist=L set block driver read-write whitelist - (affects only QEMU, not qemu-img) + (by default affects only QEMU, not tools like qemu-img) --block-drv-ro-whitelist=L set block driver read-only whitelist - (affects only QEMU, not qemu-img) + (by default affects only QEMU, not tools like qemu-img) + --enable-block-drv-whitelist-in-tools + use block whitelist also in tools instead of only QEMU --enable-trace-backends=B Set trace backend Available backends: $trace_backend_list --with-trace-file=NAME Full PATH,NAME of file to store traces
@@ -4556,6 +4563,9 @@ if test "$audio_win_int" = "yes" ; then fi echo "CONFIG_BDRV_RW_WHITELIST=$block_drv_rw_whitelist" >> $config_host_mak echo "CONFIG_BDRV_RO_WHITELIST=$block_drv_ro_whitelist" >> $config_host_mak +if test "$block_drv_whitelist_tools" = "yes" ; then + echo "CONFIG_BDRV_WHITELIST_TOOLS=y" >> $config_host_mak +fi if test "$xfs" = "yes" ; then echo "CONFIG_XFS=y" >> $config_host_mak fi
block.c +3 -0
@@ -6162,6 +6162,9 @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,void bdrv_init (void ) { +#ifdef CONFIG_BDRV_WHITELIST_TOOLS + use_bdrv_whitelist =1 ; +#endif module_call_init (MODULE_INIT_BLOCK ); }
meson.build +1 -0
@@ -2859,6 +2859,7 @@ summary_info += {'coroutine pool': config_host['CONFIG_COROUTINE_POOL'] == '1 if have_block summary_info += {'Block whitelist (rw)': config_host['CONFIG_BDRV_RW_WHITELIST']} summary_info += {'Block whitelist (ro)': config_host['CONFIG_BDRV_RO_WHITELIST']} + summary_info += {'Use block whitelist in tools': config_host.has_key('CONFIG_BDRV_WHITELIST_TOOLS')} summary_info += {'VirtFS support': have_virtfs} summary_info += {'build virtiofs daemon': have_virtiofsd} summary_info += {'Live block migration': config_host.has_key('CONFIG_LIVE_BLOCK_MIGRATION')}
Signature · 1 line
2.31.1
Quoted 7 lines · Kevin Wolf
Currently, the block driver whitelists are only applied for the system emulator. All other binaries still give unrestricted access to all block drivers. There are use cases where this made sense because the main concern was avoiding customers running VMs on less optimised block drivers and getting bad performance. Allowing the same image format e.g. as a target for 'qemu-img convert' is not a problem then.
Quoted 8 lines
However, if the concern is the supportability of the driver in general, either in full or when used read-write, not applying the list driver whitelist in tools doesn't help - especially since qemu-nbd and qemu-storage-daemon now give access to more or less the same operations in block drivers as running a system emulator.
In order to address this, introduce a new configure option that enforces the driver whitelist in all binaries.
Is it feasible that someone would want two separate lists: one for qemu (which runs run efficiently) and another for tools (which ones do we support at all)? As written, your patch offers no chance to distinguish between the two.
Also, is now a good time to join the bandwagon on picking a more descriptive name (such as 'allow-list') for this terminology?
block.c BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, · line 6166
Quoted 7 lines
+++
b/block.cblock.c +2
-0@@ -6162,6 +6162,9 @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,void bdrv_init (void ) { +#ifdef CONFIG_BDRV_WHITELIST_TOOLS + use_bdrv_whitelist =1 ;
Pre-existing, but this variable seems like a natural candidate to be bool instead of int.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Am 09.07.2021 um 19:45 hat Eric Blake geschrieben:
Quoted 14 lines · Kevin Wolf
Quoted 13 lines
Currently, the block driver whitelists are only applied for the system emulator. All other binaries still give unrestricted access to all block drivers. There are use cases where this made sense because the main concern was avoiding customers running VMs on less optimised block drivers and getting bad performance. Allowing the same image format e.g. as a target for 'qemu-img convert' is not a problem then.
However, if the concern is the supportability of the driver in general, either in full or when used read-write, not applying the list driver whitelist in tools doesn't help - especially since qemu-nbd and qemu-storage-daemon now give access to more or less the same operations in block drivers as running a system emulator.
Quoted 7 lines
Quoted 2 lines
In order to address this, introduce a new configure option that enforces the driver whitelist in all binaries.
Is it feasible that someone would want two separate lists: one for qemu (which runs run efficiently) and another for tools (which ones do we support at all)? As written, your patch offers no chance to distinguish between the two.
Possibly. However, supporting a second list would require a much larger code change than this patch, so I'd say this is a problem we should only solve when someone actually has it.
Quoted 2 lines
Also, is now a good time to join the bandwagon on picking a more descriptive name (such as 'allow-list') for this terminology?
I don't have an opinion on the time, but I do have an opinion on using a separate email thread for it. :-)
Initially I tried to find a way not to use "whitelist" in the new option name, but that only made things inconsistent and confusing, and renaming the existing options is definitely out of scope for this patch.
Quoted 10 lines
block.c BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, · line 6166
Quoted 7 lines
+++
b/block.cblock.c +2
-0@@ -6162,6 +6162,9 @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,void bdrv_init (void ) { +#ifdef CONFIG_BDRV_WHITELIST_TOOLS + use_bdrv_whitelist =1 ;Pre-existing, but this variable seems like a natural candidate to be bool instead of int.
Yes, I guess we could have a cleanup patch there.
Kevin
Eric Blake
Reviewed-byQuoted 8 lines · Kevin Wolf
Quoted 4 lines
Is it feasible that someone would want two separate lists: one for qemu (which runs run efficiently) and another for tools (which ones do we support at all)? As written, your patch offers no chance to distinguish between the two.
Possibly. However, supporting a second list would require a much larger code change than this patch, so I'd say this is a problem we should only solve when someone actually has it.
Indeed.
Quoted 6 lines
Quoted 3 lines
Also, is now a good time to join the bandwagon on picking a more descriptive name (such as 'allow-list') for this terminology?
I don't have an opinion on the time, but I do have an opinion on using a separate email thread for it. :-)
Agreed with that sentiment.
Quoted 4 lines
Initially I tried to find a way not to use "whitelist" in the new option name, but that only made things inconsistent and confusing, and renaming the existing options is definitely out of scope for this patch.
Also agreed. Therefore, reviewing this patch in isolation (even if we eventually do followups for the issues I pointed out) is fine, and you now have:
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Quoted 27 lines · Kevin Wolf
Am 09.07.2021 um 19:45 hat Eric Blake geschrieben:
Quoted 14 lines · Kevin Wolf
Quoted 13 lines
Currently, the block driver whitelists are only applied for the system emulator. All other binaries still give unrestricted access to all block drivers. There are use cases where this made sense because the main concern was avoiding customers running VMs on less optimised block drivers and getting bad performance. Allowing the same image format e.g. as a target for 'qemu-img convert' is not a problem then.
However, if the concern is the supportability of the driver in general, either in full or when used read-write, not applying the list driver whitelist in tools doesn't help - especially since qemu-nbd and qemu-storage-daemon now give access to more or less the same operations in block drivers as running a system emulator.
Quoted 7 lines
Quoted 2 lines
In order to address this, introduce a new configure option that enforces the driver whitelist in all binaries.
Is it feasible that someone would want two separate lists: one for qemu (which runs run efficiently) and another for tools (which ones do we support at all)? As written, your patch offers no chance to distinguish between the two.
Possibly. However, supporting a second list would require a much larger code change than this patch, so I'd say this is a problem we should only solve when someone actually has it.
Quoted 5 lines
Quoted 2 lines
Also, is now a good time to join the bandwagon on picking a more descriptive name (such as 'allow-list') for this terminology?
I don't have an opinion on the time, but I do have an opinion on using a separate email thread for it. :-)
It isn't difficult - the word "white" adds no value in this arg and can simply be removed entirely right now.
Quoted 3 lines
Initially I tried to find a way not to use "whitelist" in the new option name, but that only made things inconsistent and confusing, and renaming the existing options is definitely out of scope for this patch.
Calling it --block-drv-list, would be consistent with the existing --audio-drv-list. Fixing up the other existing block list args would be nice, but should not stop use of the better name in this patch right now.
Regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|