/* * netsniff-ng - the packet sniffing beast * Copyright 2011 - 2013 Daniel Borkmann. * Subject to the GPL, version 2. */ #ifndef CURVE_H #define CURVE_H #include #include #include "locking.h" #include "built_in.h" #include "xio.h" #include "crypto_box_curve25519xsalsa20poly1305.h" struct tai { uint64_t x; }; struct taia { struct tai sec; uint32_t nano; uint32_t atto; }; static struct taia tolerance_taia = { .sec.x = 0, .nano = 700000000ULL, .atto = 0, }; #define crypto_box_zerobytes crypto_box_curve25519xsalsa20poly1305_ZEROBYTES #define crypto_box_boxzerobytes crypto_box_curve25519xsalsa20poly1305_BOXZEROBYTES #define crypto_box_noncebytes crypto_box_curve25519xsalsa20poly1305_NONCEBYTES #define crypto_box_beforenmbytes crypto_box_curve25519xsalsa20poly1305_BEFORENMBYTES struct curve25519_proto { unsigned char enonce[crypto_box_noncebytes] __aligned_16; unsigned char dnonce[crypto_box_noncebytes] __aligned_16; unsigned char key[crypto_box_noncebytes] __aligned_16; }; struct curve25519_struct { unsigned char *enc_buf; unsigned char *dec_buf; size_t enc_buf_size; size_t dec_buf_size; struct spinlock enc_lock; struct spinlock dec_lock; }; extern void curve25519_selftest(void); extern void curve25519_alloc_or_maybe_die(struct curve25519_struct *curve); extern void curve25519_free(void *curve); extern int curve25519_pubkey_hexparse_32(unsigned char *bin, size_t blen, const char *ascii, size_t alen); extern int curve25519_proto_init(struct curve25519_proto *proto, unsigned char *pubkey_remote, size_t len, char *home, int server); extern ssize_t curve25519_encode(struct curve25519_struct *curve, struct curve25519_proto *proto, unsigned char *plaintext, size_t size, unsigned char **chipertext); extern ssize_t curve25519_decode(struct curve25519_struct *curve, struct curve25519_proto *proto, unsigned char *chipertext, size_t size, unsigned char **plaintext, struct taia *arrival_taia); static inline void tai_pack(unsigned char *s, struct tai *t) { uint64_t x; x = t->x; s[7] = x & 255; x >>= 8; s[6] = x & 255; x >>= 8; s[5] = x & 255; x >>= 8; s[4] = x & 255; x >>= 8; s[3] = x & 255; x >>= 8; s[2] = x & 255; x >>= 8; s[1] = x & 255; x >>= 8; s[0] = x; } static inline void tai_unpack(unsigned char *s, struct tai *t) { uint64_t x; x = (unsigned char) s[0]; x <<= 8; x += (unsigned char) s[1]; x <<= 8; x += (unsigned char) s[2]; x <<= 8; x += (unsigned char) s[3]; x <<= 8; x += (unsigned char) s[4]; x <<= 8; x += (unsigned char) s[5]; x <<= 8; x += (unsigned char) s[6]; x <<= 8; x += (unsigned char) s[7]; t->x = x; } static inline void taia_pack(unsigned char *s, struct taia *t) { unsigned long x; tai_pack(s, &t->sec); s += 8; x = t->atto; s[7] = x & 255; x >>= 8; s[6] = x & 255; x >>= 8; s[5] = x & 255; x >>= 8; s[4] = x; x = t->nano; s[3] = x & 255; x >>= 8; s[2] = x & 255; x >>= 8; s[1] = x & 255; x >>= 8; s[0] = x; } static inline void taia_unpack(unsigned char *s, struct taia *t) { unsigned long x; tai_unpack(s, &t->sec); s += 8; x = (unsigned char) s[4]; x <<= 8; x += (unsigned char) s[5]; x <<= 8; x += (unsigned char) s[6]; x <<= 8; x += (unsigned char) s[7]; t->atto = x; x = (unsigned char) s[0]; x <<= 8; x += (unsigned char) s[1]; x <<= 8; x += (unsigned char) s[2]; x <<= 8; x += (unsigned char) s[3]; t->nano = x; } #define tai_unix(t, u) ((void) ((t)->x = 4611686018427387914ULL + (uint64_t) (u))) static inline void taia_now(struct taia *t) { struct timeval now; gettimeofday(&now, NULL); tai_unix(&t->sec, now.tv_sec); t->nano = 1000 * now.tv_usec + 500; t->atto = secrand(); } static inline void taia_sub(struct taia *res, const struct taia *u, const struct taia *v) { unsigned long unano = u->nano; unsigned long uatto = u->atto; res->sec.x = u->sec.x - v->sec.x; res->nano = unano - v->nano; res->atto = uatto - v->atto; if (res->atto > uatto) { res->atto += 1000000000UL; --res->nano; } if (res->nano > unano) { res->nano += 1000000000UL; --res->sec.x; } } static inline void taia_add(struct taia *res, const struct taia *u, const struct taia *v) { res->sec.x = u->sec.x + v->sec.x; res->nano = u->nano + v->nano; res->atto = u->atto + v->atto; if (res->atto > 999999999UL) { res->atto -= 1000000000UL; ++res->nano; } if (res->nano > 999999999UL) { res->nano -= 1000000000UL; ++res->sec.x; } } static inline int taia_less(const struct taia *t, const struct taia *u) { if (t->sec.x < u->sec.x) return 1; if (t->sec.x > u->sec.x) return 0; if (t->nano < u->nano) return 1; if (t->nano > u->nano) return 0; return t->atto < u->atto; } static inline int is_good_taia(struct taia *arrival_taia, struct taia *packet_taia) { int is_ts_good = 0; struct taia sub_res; if (taia_less(arrival_taia, packet_taia)) { taia_sub(&sub_res, packet_taia, arrival_taia); if (taia_less(&sub_res, &tolerance_taia)) is_ts_good = 1; else is_ts_good = 0; } else { taia_sub(&sub_res, arrival_taia, packet_taia); if (taia_less(&sub_res, &tolerance_taia)) is_ts_good = 1; else is_ts_good = 0; } return is_ts_good; } #endif /* CURVE_H */ a href='/cgit.cgi/linux/net-next.git/diff/?h=packet-loop-back&id=6d037de90a1fd7b4879b48d4dd5c4839b271be98&id2=4c2e07c6a29e0129e975727b9f57eede813eea85'>diff)
MIPS: Fix possible corruption of cache mode by mprotect.
The following testcase may result in a page table entries with a invalid CCA field being generated: static void *bindstack; static int sysrqfd; static void protect_low(int protect) { mprotect(bindstack, BINDSTACK_SIZE, protect); } static void sigbus_handler(int signal, siginfo_t * info, void *context) { void *addr = info->si_addr; write(sysrqfd, "x", 1); printf("sigbus, fault address %p (should not happen, but might)\n", addr); abort(); } static void run_bind_test(void) { unsigned int *p = bindstack; p[0] = 0xf001f001; write(sysrqfd, "x", 1); /* Set trap on access to p[0] */ protect_low(PROT_NONE); write(sysrqfd, "x", 1); /* Clear trap on access to p[0] */ protect_low(PROT_READ | PROT_WRITE | PROT_EXEC); write(sysrqfd, "x", 1); /* Check the contents of p[0] */ if (p[0] != 0xf001f001) { write(sysrqfd, "x", 1); /* Reached, but shouldn't be */ printf("badness, shouldn't happen but does\n"); abort(); } } int main(void) { struct sigaction sa; sysrqfd = open("/proc/sysrq-trigger", O_WRONLY); if (sigprocmask(SIG_BLOCK, NULL, &sa.sa_mask)) { perror("sigprocmask"); return 0; } sa.sa_sigaction = sigbus_handler; sa.sa_flags = SA_SIGINFO | SA_NODEFER | SA_RESTART; if (sigaction(SIGBUS, &sa, NULL)) { perror("sigaction"); return 0; } bindstack = mmap(NULL, BINDSTACK_SIZE, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); if (bindstack == MAP_FAILED) { perror("mmap bindstack"); return 0; } printf("bindstack: %p\n", bindstack); run_bind_test(); printf("done\n"); return 0; } There are multiple ingredients for this: 1) PAGE_NONE is defined to _CACHE_CACHABLE_NONCOHERENT, which is CCA 3 on all platforms except SB1 where it's CCA 5. 2) _page_cachable_default must have bits set which are not set _CACHE_CACHABLE_NONCOHERENT. 3) Either the defective version of pte_modify for XPA or the standard version must be in used. However pte_modify for the 36 bit address space support is no affected. In that case additional bits in the final CCA mode may generate an invalid value for the CCA field. On the R10000 system where this was tracked down for example a CCA 7 has been observed, which is Uncached Accelerated. Fixed by: 1) Using the proper CCA mode for PAGE_NONE just like for all the other PAGE_* pte/pmd bits. 2) Fix the two affected variants of pte_modify. Further code inspection also shows the same issue to exist in pmd_modify which would affect huge page systems. Issue in pte_modify tracked down by Alastair Bridgewater, PAGE_NONE and pmd_modify issue found by me. The history of this goes back beyond Linus' git history. Chris Dearman's commit 351336929ccf222ae38ff0cb7a8dd5fd5c6236a0 ("[MIPS] Allow setting of the cache attribute at run time.") missed the opportunity to fix this but it was originally introduced in lmo commit d523832cf12007b3242e50bb77d0c9e63e0b6518 ("Missing from last commit.") and 32cc38229ac7538f2346918a09e75413e8861f87 ("New configuration option CONFIG_MIPS_UNCACHED.") Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Reported-by: Alastair Bridgewater <alastair.bridgewater@gmail.com>