random: fix priming of last_data
Subsystems (3)
PATCH 1/1 random: fix priming of last_data
Commit ec8f02da9e added priming of last_data per fips requirements. Unfortuantely, it did so in a way that can lead to multiple threads all incrementing nbytes, but only one actually doing anything with the extra data, which leads to some fun random corruption and panics.
The fix is to simply do everything needed to prime last_data in a single shot, so there's no window for multiple cpus to increment nbytes -- in fact, we won't even increment or decrement nbytes anymore, we'll just extract the needed EXTRACT_BYTES one time per pool and then carry on with the normal routine.
All these changes have been tested across multiple hosts and architectures where panics were previously encoutered. The code changes are are strictly limited to areas only touched when when booted in fips mode.
This change should also go into 3.8-stable, to make the myriads of fips users on 3.8.x happy.
--- drivers/char/random.c | 30 +++++++++++++++--------------- 1 files changed, 15 insertions(+), 15 deletions(-)
drivers/char/random.c +15 -15
@@ -953,10 +953,23 @@ static ssize_t extract_entropy(struct entropy_store *r, void *buf, {ssize_t ret =0 , i;__u8 tmp[EXTRACT_SIZE ]; +unsigned long flags;/* if last_data isn't primed, we need EXTRACT_SIZE extra bytes */ -if (fips_enabled && !r->last_data_init) - nbytes +=EXTRACT_SIZE ; +if (fips_enabled) { +spin_lock_irqsave (&r->lock, flags); +if (!r->last_data_init) { + r->last_data_init = true; +spin_unlock_irqrestore (&r->lock, flags); +trace_extract_entropy (r->name,EXTRACT_SIZE , + r->entropy_count, _RET_IP_); +xfer_secondary_pool (r,EXTRACT_SIZE ); +extract_buf (r, tmp); +spin_lock_irqsave (&r->lock, flags); +memcpy (r->last_data, tmp,EXTRACT_SIZE ); + } +spin_unlock_irqrestore (&r->lock, flags); + }trace_extract_entropy (r->name, nbytes, r->entropy_count, _RET_IP_);xfer_secondary_pool (r, nbytes);
@@ -966,19 +979,6 @@ static ssize_t extract_entropy(struct entropy_store *r, void *buf,extract_buf (r, tmp);if (fips_enabled) { -unsigned long flags; - - -/* prime last_data value if need be, per fips 140-2 */ -if (!r->last_data_init) { -spin_lock_irqsave (&r->lock, flags); -memcpy (r->last_data, tmp,EXTRACT_SIZE ); - r->last_data_init = true; - nbytes -=EXTRACT_SIZE ; -spin_unlock_irqrestore (&r->lock, flags); -extract_buf (r, tmp); - } -spin_lock_irqsave (&r->lock, flags);if (!memcmp (tmp, r->last_data,EXTRACT_SIZE ))panic ("Hardware RNG duplicated output!\n" );
Signature · 1 line
1.7.1
Quoted 9 lines · Jarod Wilson
Commit
ec8f02da9eadded priming oflast_dataper fips requirements. Unfortuantely, it did so in a way that can lead to multiple threads all incrementing nbytes, but only one actually doing anything with the extra data, which leads to some fun random corruption and panics.The fix is to simply do everything needed to prime
last_datain a single shot, so there's no window for multiple cpus to increment nbytes -- in fact, we won't even increment or decrement nbytes anymore, we'll just extract the neededEXTRACT_BYTESone time per pool and then carry on with
^^^^^ EXTRACT_SIZE
Error in the description, if anyone actually cares. Oops.
Quoted 8 lines
the normal routine.
All these changes have been tested across multiple hosts and architectures where panics were previously encoutered. The code changes are are strictly limited to areas only touched when when booted in fips mode.
This change should also go into 3.8-stable, to make the myriads of fips users on 3.8.x happy.
Signature · 2 lines
Jarod Wilson [email protected]
Quoted 86 lines · Jarod Wilson
Commit
ec8f02da9eadded priming oflast_dataper fips requirements. Unfortuantely, it did so in a way that can lead to multiple threads all incrementing nbytes, but only one actually doing anything with the extra data, which leads to some fun random corruption and panics.The fix is to simply do everything needed to prime
last_datain a single shot, so there's no window for multiple cpus to increment nbytes -- in fact, we won't even increment or decrement nbytes anymore, we'll just extract the neededEXTRACT_BYTESone time per pool and then carry on with the normal routine.All these changes have been tested across multiple hosts and architectures where panics were previously encoutered. The code changes are are strictly limited to areas only touched when when booted in fips mode.
This change should also go into 3.8-stable, to make the myriads of fips users on 3.8.x happy.
Tested-by Jan Stancek <[email protected]>Tested-by Jan Stodola <[email protected]>CC Herbert Xu <[email protected]>CC Neil Horman <[email protected]>CC "David S. Miller" <[email protected]>CC Matt Mackall <[email protected]>CC "Theodore Ts'o" <[email protected]>Signed-off-by Jarod Wilson <[email protected]>--- drivers/char/random.c | 30 +++++++++++++++--------------- 1 files changed, 15 insertions(+), 15 deletions(-)drivers/char/random.c +15
-15@@ -953,10 +953,23 @@ static ssize_t extract_entropy(struct entropy_store *r, void *buf, {ssize_t ret =0 , i;__u8 tmp[EXTRACT_SIZE ]; +unsigned long flags;/* if last_data isn't primed, we need EXTRACT_SIZE extra bytes */ -if (fips_enabled && !r->last_data_init) - nbytes +=EXTRACT_SIZE ; +if (fips_enabled) { +spin_lock_irqsave (&r->lock, flags); +if (!r->last_data_init) { + r->last_data_init = true; +spin_unlock_irqrestore (&r->lock, flags); +trace_extract_entropy (r->name,EXTRACT_SIZE , + r->entropy_count, _RET_IP_); +xfer_secondary_pool (r,EXTRACT_SIZE ); +extract_buf (r, tmp); +spin_lock_irqsave (&r->lock, flags); +memcpy (r->last_data, tmp,EXTRACT_SIZE ); + } +spin_unlock_irqrestore (&r->lock, flags); + }trace_extract_entropy (r->name, nbytes, r->entropy_count, _RET_IP_);xfer_secondary_pool (r, nbytes);@@ -966,19 +979,6 @@ static ssize_t extract_entropy(struct entropy_store *r, void *buf,extract_buf (r, tmp);if (fips_enabled) { -unsigned long flags; - - -/* prime last_data value if need be, per fips 140-2 */ -if (!r->last_data_init) { -spin_lock_irqsave (&r->lock, flags); -memcpy (r->last_data, tmp,EXTRACT_SIZE ); - r->last_data_init = true; - nbytes -=EXTRACT_SIZE ; -spin_unlock_irqrestore (&r->lock, flags); -extract_buf (r, tmp); - } -spin_lock_irqsave (&r->lock, flags);if (!memcmp (tmp, r->last_data,EXTRACT_SIZE ))panic ("Hardware RNG duplicated output!\n" );Signature · 1 line
1.7.1